What Is Earnings at Risk?

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-14 · Updated 2026-10-03

Earnings at risk asks a question a balance sheet measure cannot: not what the holdings are worth, but how much income could fail to arrive over the next year.

How it works

Specify a horizon and a confidence level — typically twelve months and something like 95% — and model how earnings respond to the relevant variables.

The output is a single figure. Earnings will not fall more than this amount, on this proportion of modeled outcomes, over this period.

It targets the income statement. Value at risk describes what a portfolio’s value could do; this describes what the reported profit could do, and the two can point differently.

Why the distinction is not academic

Earnings determine covenants, dividends and credibility. A company whose assets are fine and whose earnings miss faces breached covenants, a cut dividend and a share price reaction.

And banks are the classic case. A rate move changes the value of the bond portfolio and separately changes net interest income, and the two effects can have opposite signs.

So a bank managing only one of them is managing half its exposure, which is why regulators require both an economic value measure and an earnings measure.

A worked example

A bank funds long-dated fixed-rate loans with short-term deposits. Rates rise sharply.

The loan book’s value falls, which a value-based measure captures immediately and which matters only if the loans are sold.

Deposit costs reprice upward while loan income does not. Net interest income compresses, and that shows up in every quarterly report until the loans mature or reprice.

Earnings at risk quantifies the second effect. It answers how much of next year’s income that repricing could remove, which is the number the board and the covenant actually respond to.

What drives the answer

The rate scenario. A parallel shift, a steepening, a flattening and an inversion produce entirely different results from the same balance sheet.

Deposit behavior assumptions. How quickly savers move money when rates rise is the single largest assumption in most bank models, and it is estimated from history that may not repeat.

Volume assumptions. Whether the balance sheet is held constant or allowed to grow changes the figure substantially, and both conventions are used.

Which means two institutions can report very different figures from similar positions. The measure is only as meaningful as its stated assumptions, and comparing headline numbers across firms without reading those assumptions compares modeling choices rather than risk.

The original data

Measured on the S&P 500 since 1950, 93% of trading days ended below the previous record, the worst fall was 56.8%, from 2007 to 2009, and the longest below-peak stretch ran 7.5 years, from 1973 to 1980, before new records followed.

That 73-bar stretch is what an earnings measure has to span. A one-year horizon covers a period long enough for a sustained adverse move, and short enough that a recovery beyond it does not help the reported numbers.

And this site’s fee measurement is the compounding comparison: 75 basis points costs 20.2% of a thirty-year balance. A margin compression of similar size, sustained across a lending book, is the same arithmetic applied to a business’s income.

Where it is used beyond banking

Utilities and energy companies. Earnings depend on commodity prices and the measure is applied to fuel costs and power prices directly.

Any exporter. Revenue in one currency and costs in another means earnings vary with the exchange rate, independently of how the business performs.

And corporate treasuries generally. Anybody hedging an exposure needs to know how much of next year’s income is at stake, because that determines how much hedging is worth its cost.

The common thread is that it answers a budgeting question. Value measures serve investors assessing worth; earnings measures serve managers deciding how much certainty to buy — and those are different audiences asking different things.

How it differs from value at risk

The horizon is longer. Value at risk is typically a one-day or ten-day measure; earnings at risk runs to a quarter or a year, because that is the reporting period being protected.

The accounting treatment matters. Whether an instrument is marked to market or held at amortized cost changes whether a move hits earnings at all, so two economically identical positions can have very different earnings-at-risk figures.

And balance sheet changes must be modeled. A one-day measure can hold positions constant; a one-year measure cannot, because loans mature, deposits reprice and new business is written.

Which makes it the harder calculation of the two. It requires assumptions about behavior and volumes that a short-horizon value measure can ignore entirely.

How this connects

The rate scenarios are shapes of the yield curve. A steepening or an inversion changes the gap between what a bank pays on deposits and earns on loans, which is why curve shape matters more than the level alone.

The economic value measure it is paired with rests on duration, the sensitivity of the bank’s assets and liabilities to a change in rates.

The usual tool for reducing the figure is an interest rate swap, exchanging fixed payments for floating ones so income reprices in step with funding costs.

Exporters narrow the currency version with a forward contract, fixing the conversion rate for revenue that has not yet arrived.

Liquidity risk is the close neighbor in the failure described below. Deposits leaving quickly is a funding problem as well as an earnings problem, and the two arrive together.

Credit risk is usually modeled separately, although a recession that cuts rates also raises loan losses, so the two measures can understate a combined hit.

Confidence-level measures built from standard deviation describe ordinary outcomes well, and tail risk sits precisely where they stop describing anything.

And operational risk sits outside the model entirely, even though a failed system or fraud reduces earnings just as surely.

When it fails

The characteristic failure is a deposit assumption that stops holding. A bank models that savers move slowly when rates rise, because they historically have, and its earnings-at-risk figure is comfortable. Then rates rise in an era of instant transfers and rate-comparison apps, deposits leave far faster than any historical episode, and funding costs reprice almost immediately while the loan book does not. The model was correctly built on the available history, and the behavior it described belonged to a period before moving money took thirty seconds.

A second failure is modeling only a parallel rate shift, when a flattening curve is what actually compresses a lending margin.

A third is comparing figures across institutions without matching the assumptions behind them.

A fourth is treating the confidence level as a limit. Outcomes beyond it exist and are not described.

The question: The model says the income holds. Does it?

And a fifth is managing earnings while ignoring economic value, which is how a position can look fine on one measure and be badly wrong on the other.

Related

Market risk covers the exposure most of these models quantify. Business risk covers the wider set of things that damage earnings. And cash-flow-to-debt ratio covers what a squeezed income does to debt capacity.

This page is educational, not financial advice. Test every idea on your own charts before risking money.