Financial Risk Management

Recognising the different sources of risk

Market risk arises when movements in prices or rates change the value of a position. Equity prices, interest rates, exchange rates and commodity prices are distinct market factors, but a single transaction can expose a business to several of them. A dollar-denominated bond held by a euro-based investor has interest-rate exposure and currency exposure; if the issuer can default, it also has credit exposure. Describing it simply as a bond does not identify the risks that determine its value.

Credit risk concerns failure to receive contractual payments and deterioration in a counterparty's ability to pay. Default probability matters, but so do the amount owed at default, the recovery available, concentration in particular borrowers or industries, and country conditions that can impede repayment. Liquidating collateral may take time and yield less than its recorded value. A large number of loans can still constitute a concentrated portfolio when borrowers depend on the same economic conditions.

Liquidity risk has two connected dimensions. Funding liquidity is the ability to meet payments and refinance obligations when due. Market liquidity is the ability to sell or hedge an asset without an excessive price concession. A solvent-looking institution can fail if it cannot raise cash quickly enough. Conversely, forced sales made to obtain cash can crystallise losses and weaken capital. The distinction between a loss of value and a shortage of immediately available money is therefore fundamental.

Operational risk includes failures of processes, people, technology and internal controls, as well as external events. Legal and compliance failures can produce fines, compensation obligations and restrictions on business. Strategic or business risk concerns the viability of the business model, competitive position and choices about products and markets. Reputational damage can amplify any of these risks by undermining customer, investor or counterparty confidence. The categories organise analysis; they do not imply that actual losses respect organisational boundaries.

Systemic risk concerns disruption of the financial system rather than only the failure of one institution. Common exposures, interconnected claims and simultaneous attempts to sell assets can transmit problems across firms. Model risk arises when decisions rely on an inappropriate model, unreliable inputs or assumptions that no longer describe the environment. Incentive risk arises when a decision-maker benefits from taking risks whose costs fall elsewhere. A distressed firm's shareholders may prefer a gamble that offers a chance of recovery while transferring much of the downside to creditors, deposit insurers or taxpayers. This is the logic of gambling for resurrection.

The risk-management cycle and the choice of response

A useful risk-management process begins with identification: what could go wrong, through which mechanism and for whom? Internal incidents, external failures and structured self-assessment help reveal exposures that are not apparent from a balance sheet. Assessment then connects causes to financial consequences, identifies the amount exposed, and estimates the frequency and severity of losses where quantification is credible. A rare loss capable of threatening survival deserves attention even if its expected annual cost looks small.

The institution next chooses controls and limits consistent with its risk appetite. An exposure limit constrains the amount at risk; a concentration limit prevents dependence on a single source of repayment; a loss or capital limit constrains the consequences of adverse scenarios. Finally, monitoring tests whether the controls work, whether limits are respected and whether the original assumptions remain valid. New information should feed back into identification and assessment. A policy that is written once and never challenged is not a complete risk-management system.

Risk can be retained when its size is acceptable and the institution can absorb it. Retention should be a conscious decision rather than the accidental result of failing to measure an exposure. Avoidance removes the activity generating the risk, for example by leaving a market or closing a business line. Reduction changes the probability or severity of a loss: better information systems reduce processing failures, while collateral may reduce the loss associated with a borrower's default.

Risk transfer reallocates consequences through insurance, contractual indemnities, outsourcing or other agreements. Outsourcing a process does not remove responsibility for overseeing it, and insurance creates its own exclusions, limits and counterparty dependence. Hedging uses offsetting financial positions, such as futures, options or credit derivatives. It can reduce the sensitivity of a portfolio to a specified risk factor while introducing basis risk, collateral needs or counterparty exposure. The relevant question is how the combined position behaves, not whether the hedge itself ever loses money.

A risk manager need not personally program every calculation, but must understand what the model measures, specify the analysis required and interpret its limitations. Specialist software and AI tools can assist with calculations and explanations, yet their output still needs checking against definitions, units, known examples and reliable sources. A plausible-looking formula or chart can be wrong. Independent reproduction of a material result is particularly valuable when the conclusion affects capital, limits or investment decisions.

Governance, independent control and the three lines

The three-lines framework assigns different responsibilities to the people who take risk, those who challenge it and those who provide independent assurance. The first line consists of operating management and the business units that originate transactions. They own the risks of their activities and must maintain effective day-to-day controls. The second line includes risk management and compliance functions that establish frameworks, monitor exposures and challenge decisions. These functions require sufficient authority, expertise and access to information to disagree with revenue-generating teams.

Internal audit forms the third line. It assesses whether governance, risk management and controls are designed and operating effectively, and reports independently to the governing body or its audit committee. The board oversees strategy and risk appetite; senior management implements them. External audit and supervision supply additional scrutiny, but neither replaces the institution's own lines of responsibility. A diagram with three boxes is ineffective if staff cannot escalate concerns or if managers can alter the information presented to the board.

Independence matters especially when performance is rewarded through short-term profits. A trader can earn an attractive return by selling protection against an infrequent catastrophe: profits appear stable until the adverse event occurs. If the trader also determines the reported risk and can change loss scenarios, both the apparent return and the apparent risk may be misleading. Controls should therefore protect valuation, model validation and communication of risk information from manipulation by the people whose performance is being assessed.

Effective oversight also requires understanding the economic purpose of a transaction. A derivative loss can be consistent with a successful hedge if the underlying exposure gains value. Equally, a profitable derivative position can conceal a dangerous increase in unhedged risk. Boards need reports that combine exposures, hedges, funding requirements and stress outcomes, with clear explanations of assumptions and exceptions.

Legal reference: German Banking Act, section 54a.

Catastrophe bonds and the distinction between an event and a valuation loss

Catastrophe bonds illustrate why risk knowledge matters beyond trading desks and banks. These securities transfer specified disaster losses to investors. Their terms define triggers, such as a particular insured loss, an industry loss measure or a physical characteristic of a hurricane or earthquake. Investors receive compensation for accepting the possibility that interest or principal will be reduced when the contractual trigger is met. The exact trigger and the assets securing the obligation must be understood before interpreting the coupon as an investment return.

The absence of a trigger does not guarantee a stable market price. Investors may revise the estimated probability of a disaster, demand a higher risk premium, reassess the quality of catastrophe models or need to sell into an illiquid market. A bond can consequently lose market value without yet suffering a contractual principal write-down. An institution investing money needed for predictable future commitments must consider this valuation and liquidity risk as well as the trigger probability.

Open figure at full size ↗

The same distinction applies to apparently safe banking assets. A government bond may retain a high probability of repayment at maturity while losing substantial market value when interest rates rise. Whether the holder can wait until maturity depends on its liabilities and available liquidity. A promise of eventual repayment does not solve an immediate funding problem.

A banking failure need not appear as an ordinary corporate bankruptcy. A bank may be resolved, sold, merged or supported through an institutional protection arrangement. A June 2025 report on Raiffeisenbank im Hochtaunus described an expected EUR 1.2 billion support burden for the cooperative banking protection scheme and identified it as a further case within that sector. The relevant lesson is that intervention can prevent a disorderly closure while still revealing serious losses and weaknesses in the failed institution's risk management.

Open figure at full size ↗

Silicon Valley Bank: interest-rate risk, liquidity and incentives

Silicon Valley Bank's March 2023 failure demonstrates how several risks reinforce one another. Rapid deposit growth financed a large portfolio of longer-duration securities. When interest rates rose, the market value of those securities fell. The bank also relied heavily on concentrated, uninsured deposits whose owners could withdraw funds quickly. A portfolio that appears manageable under an assumption of stable funding becomes much more vulnerable when funding leaves and assets must be monetised.

The Federal Reserve's 2023 review found weaknesses in governance, liquidity planning and interest-rate-risk management. It described management's focus on short-run earnings, breaches of longer-term interest-rate-risk limits and the removal during 2022 of hedges that would have helped protect against rising rates. Realising a gain on a hedge does not create protection against the next adverse movement: after the position is closed, the exposure it offset remains unless another adjustment replaces it.

Compensation was part of this governance problem. Supervisors found that incentive decisions relied heavily on financial performance and gave insufficient weight to risk-management deficiencies. This supports a lesson about incentives and accountability. It does not establish the stronger claim that particular managers removed particular swaps solely to obtain a specified bonus, or that bonuses alone caused the failure. A careful explanation distinguishes documented weaknesses from an inferred personal motive.

The immediate crisis involved a loss of depositor confidence, a securities sale that crystallised losses and exceptionally rapid withdrawals. The bank was closed on 10 March 2023. The broader lesson is that interest-rate risk, funding concentration, unrealised losses and inadequate contingency arrangements cannot be analysed in isolation. Accounting classifications may affect when losses appear in reported earnings, but they do not remove the economic loss or the cash needed to meet withdrawals.

Primary sources: Federal Reserve review of SVB, risk management (2023); Federal Reserve review of incentive compensation (2023).

Investment mandates, misconduct and supervisory intervention

The Berlin dentists' pension institution, VZB, provides a different governance setting. Retirement promises require investment decisions consistent with liquidity needs, diversification and the institution's mandate. Contemporary reporting described very large losses and disputed exposure to risky, illiquid investments. Claims against former managers, advisers, auditors or supervisory bodies concern different duties and require separate evidence. A risk report produced by an outside provider does not, by itself, establish that the governing body discharged its own responsibilities.

The Berlin Labour Court confirmed in July 2026 that VZB sought approximately EUR 50 million in damages from its former director and that procedural questions remained before the court. That is a confirmed claim, not a final finding of liability. Reported portfolio percentages, total loss estimates and the number of people under investigation should not be treated as settled judicial facts. The study lesson is the danger of allowing an investment mandate, actual portfolio risk and independent oversight to diverge, together with the importance of distinguishing allegations, valuations and adjudicated responsibility.

Open figure at full size ↗

The Allianz Global Investors US Structured Alpha case involved a documented fraud in which investors were misled about downside protection and risk. Altered risk information and misrepresented hedging practices prevented investors from understanding the exposures they accepted. Following a guilty plea, the July 2023 sentence included more than USD 463 million in forfeiture, USD 3.23 billion in restitution and USD 2.33 billion in fines. Describing the entire resolution as a single fine for breaching a three-lines diagram is inaccurate: the underlying misconduct and the categories of financial sanction matter.

N26 illustrates how compliance weaknesses can restrict business growth. BaFin imposed measures relating to anti-money-laundering controls, appointed special monitoring and restricted customer acquisition. The historical growth restriction began in 2021 and was lifted effective June 2024, according to N26's announcement. Separate fines of EUR 4.25 million in 2021 and EUR 9.2 million announced in 2024 concerned delayed suspicious-activity reporting. The economic cost of deficient controls includes remediation expenditure and lost business opportunities, as well as the fine itself.

German banking law also makes specified failures of risk governance potentially relevant to criminal responsibility. Section 54a KWG provides for imprisonment of up to five years or a fine in the circumstances it defines. The offence requires more than an investment loss: it connects specified management duties, a binding supervisory remediation order and an ensuing threat to the institution's continued existence. Negligently causing the specified danger is treated separately. Risk management is therefore both an economic discipline and a set of responsibilities whose legal consequences depend on the facts and applicable provisions.

Case documentation: US Department of Justice, Structured Alpha sentencing (2023); Berlin Labour Court, pension-fund proceedings (2026); BaFin, N26 restrictions (2021); N26, fine disclosure (2024); N26, removal of the growth restriction (2024).

Recurring lessons from historical failures

Financial history contains repeated examples of familiar exposures interacting with leverage, liquidity and weak controls. The US savings-and-loan crisis exposed the vulnerability of institutions funding long-term assets with liabilities whose cost could rise much faster. Metallgesellschaft's 1993 oil-related difficulties showed that a hedge can generate severe intermediate cash requirements when the maturity and settlement pattern of derivatives differ from the commercial contracts being hedged. The lesson is not that all hedging is dangerous, but that funding the hedge is part of its design.

Long-Term Capital Management in 1998 combined leveraged positions with assumptions about market relationships and liquidity that became unreliable under stress. The episodes involving Nick Leeson at Barings in 1995 and Jérôme Kerviel at Société Générale in 2008 illustrate the importance of trade verification, position limits and separation of duties. JPMorgan's 2012 London Whale losses involved complex credit derivatives and weaknesses in oversight and risk reporting. A model's numerical output cannot substitute for a clear understanding of the position it describes.

Lehman Brothers' failure in 2008 intensified an already developing global financial crisis; it did not single-handedly initiate every aspect of that crisis. Exposure to mortgage-related assets, leverage and fragile funding were central concerns. Cum-ex transactions, in contrast, illustrate legal and conduct risk arising from arrangements associated with improper tax refunds. These cases should not be reduced to one interchangeable story about volatility. Their common feature is that a narrow view of reported profit or a single risk metric failed to capture the full consequences of the activity.

Loss distributions, Value at Risk and Expected Shortfall

Let $L$ denote the monetary loss over a specified horizon, with positive values representing losses and negative values representing gains. A risk measure summarises some feature of its distribution. Variance measures dispersion around the mean; it treats favourable and adverse deviations symmetrically. Maximum possible loss asks for the worst outcome permitted by the position's payoff. It can be limited for an unleveraged long investment, but unbounded in an idealised model for certain short positions. Neither quantity is interchangeable with a high-confidence loss quantile.

Value at Risk at confidence level $\alpha$ is the loss quantile:

A one-day 99% VaR of USD 1 million means that the model assigns approximately 1% probability to a loss above USD 1 million over one day. It is not the maximum loss that can occur. The horizon, confidence level, valuation convention and assumptions about the portfolio must accompany the number. A 95% annual VaR and a 99% daily VaR answer different questions and cannot be compared simply by looking at their monetary amounts.

Expected Shortfall, also called ES or CVaR, measures the average loss in the worst tail. A definition that also handles distributions with discrete probability masses is:

For a continuous loss distribution, this is the expected loss conditional on being beyond the VaR threshold. ES distinguishes portfolios with the same VaR but very different losses in extreme scenarios. It also requires information about a part of the distribution where data are scarce and modelling assumptions are consequential. A precise-looking tail estimate should therefore be read together with stress tests and model limitations.

A stress test asks how a portfolio or institution performs under a specified adverse scenario: for example, a recession combined with higher defaults, falling collateral values and stressed funding. It need not assign that scenario a precise probability. A supervisory stress result showing a reduced capital ratio is conditional on the scenario and methodology, not a forecast that the bank will fail with that ratio. Published rankings are meaningful only after the starting position, projected losses and scenario assumptions are understood.

For example, LBBW's published 2025 exercise reported a transitional CET1 ratio of 6.82% at the end of the adverse 2027 scenario, compared with a restated starting ratio of 16.85%. Its fully loaded adverse ratio was 5.54%, using a different regulatory basis. These are scenario outcomes under specified capital rules, not alternative observations of an actual future capital ratio. Mixing transitional and fully loaded figures, or ignoring the restatement of the starting position, produces misleading comparisons.

Maximum drawdown supplies a complementary path-based measure. If $V_t$ is portfolio value and $H_t={max}_{s\le t}{V}_s$ is its previous peak, percentage drawdown is $1-V_t/H_t$ and maximum drawdown is the largest such decline over the observed interval. It depends on the sequence of values, unlike a single terminal-return quantile. Historical maximum drawdown is the worst peak-to-trough decline in that history, not a bound on the losses that could occur in the future.

Expected loss, unexpected loss and economic capital

For a simple credit exposure, expected loss is commonly represented as:

PD is the probability of default over the chosen horizon, LGD is the fraction of exposure lost conditional on default, and EAD is the monetary exposure at default. The expression is exact when LGD and EAD are fixed conditional inputs. With stochastic exposures and recoveries, dependence must be handled consistently rather than multiplying unrelated unconditional averages. At portfolio level, expected loss is $E\left[L\right]$.

Unexpected loss is used in more than one way. Here, $UL=\sigma \left(L\right)$ denotes the standard deviation of portfolio credit losses. It measures dispersion around expected loss. Economic capital is a separate tail-based quantity: the buffer intended to absorb losses above the expected amount up to a selected confidence level:

At 99.95% confidence, the model assigns 0.05% probability to losses beyond the selected threshold. That percentage is not the standard deviation and should not be called an amount of unexpected loss without specifying the convention. Economic capital is also a modelling and management concept; it need not equal regulatory capital.

Expected credit losses should be incorporated into lending prices, provisioning and the assessment of profitability. Capital is then intended to absorb adverse deviations beyond the loss already expected and allowed for. This is the economic logic behind saying that lending customers pay for expected loss through risk-adjusted interest rates. It is not a guarantee that expected losses have already been collected in cash or fully provisioned. Underpricing, unexpected changes in default risk and accounting timing can leave a real shortfall.

Risk-weighted assets and the role of regulatory capital

Risk-weighted assets, or RWA, translate different regulated exposures into a common denominator for capital ratios. Credit, market and operational risks contribute through their respective regulatory calculations; off-balance-sheet commitments and derivatives cannot simply be omitted because they are not ordinary loans on the balance sheet. The Common Equity Tier 1 ratio is:

CET1 is a defined regulatory category of high-quality capital after relevant adjustments, not every item described as equity in the accounts. Minimum ratios, buffers and additional requirements determine how the resulting ratio is assessed. RWA is not a forecast of monetary losses and is not identical to total assets. A lower measured RWA can reflect changes in asset composition or regulatory treatment without a proportional reduction in every economic risk.

For credit risk, a standardised approach applies prescribed treatments that may depend on exposure type and eligible external ratings. Internal-ratings-based approaches use permitted bank estimates within a supervisory framework and formula. The following historical rating illustration shows the logic of assigning different weights; it is not a current capital-calculation schedule. The precise rating boundaries, exposure definitions, national implementation and applicable version must be checked before using regulatory weights in practice.

The same historical illustration assigned 35% to the stated residential category, 100% to the commercial category and 75% to other qualifying retail exposures. These figures illustrate the use of exposure classes rather than a universal principle that every residential asset is always weighted at 35%. The heading “unrated” does not mean that all retail exposures fall into a single unrestricted treatment.

A legacy corporate IRB expression makes the relationship between expected loss and capital more explicit. Writing $N$ for the standard-normal cumulative distribution and $G=N^{-1}$ for its inverse, a historical specification is:

Here $K$ is a capital requirement per unit of exposure, $R$ is the prescribed asset-correlation parameter, $M$ is effective maturity in years and $b$ is the prescribed maturity-adjustment function. PD and LGD retain their credit-risk meanings. The bracket measures a stressed conditional loss less expected loss; the final fraction adjusts maturity. The corresponding risk-weight conversion is $RW=12.5K$, so $RWA=12.5KEAD$ in this representation. The 12.5 factor is the reciprocal of 8%. The historical factor 1.06 was removed in the finalised Basel III framework; carrying it mechanically into a current calculation would be incorrect. This introduction explains the structure, while detailed credit-risk modelling requires the applicable parameters and rules.

Regulatory and historical references: Basel III reform summary; LBBW, 2025 stress-test results.

Risk, uncertainty and the shape of distributions

Risk is often used for situations in which probabilities can be estimated with some credibility; uncertainty describes situations in which the relevant mechanisms or probabilities are not reliably known. In the tradition associated with Frank Knight, the distinction matters because adding decimal places to a model does not resolve ignorance about the process generating losses. Familiar credit defaults may be represented by a statistical model. A new geopolitical disruption may require scenarios with uncertain probabilities. An unrecognised failure mechanism calls for resilience and contingency capacity as well as measurement.

The mean is the expected value, $\mu =E\left[X\right]$. Variance is the second central moment, ${\sigma }^2=E\left[{(X-\mu )}^2\right]$, and standard deviation is its square root. In financial applications, volatility usually means the standard deviation of returns over a specified horizon or its annualised counterpart. It is essential to state the return definition, frequency and estimation period: daily simple-return volatility is not automatically interchangeable with annual log-return volatility.

Skewness and kurtosis describe features that variance alone misses:

Negative skewness indicates a longer or more influential left tail, while positive skewness indicates the opposite direction. When low returns represent losses, negative skewness warns that adverse outcomes may be more severe than a symmetric approximation suggests. However, skewness alone does not mathematically determine every VaR quantile: the complete distributions and the confidence level matter.

A normal distribution has kurtosis 3, or excess kurtosis zero when 3 is subtracted. Kurtosis above 3 signals a larger standardised fourth moment and often accompanies more extreme observations than a normal model predicts. It is a useful warning about tail-sensitive modelling, but is not a complete description of the tail or a universal ranking of all extreme quantiles. Some heavy-tailed distributions do not even possess a finite fourth moment. Stress analysis is especially valuable when historical data are too sparse to estimate such features reliably.

Portfolio return, covariance and diversification

Markowitz portfolio analysis evaluates combinations of expected return and variance. A portfolio lies on the efficient frontier when no feasible alternative offers a higher expected return for the same risk, or a lower risk for the same expected return. Risk aversion does not imply selecting the lowest-volatility asset in isolation. It means choosing an appropriate efficient combination, given the investor's objective and constraints.

Open figure at full size ↗

Open figure at full size ↗

Let $w_i$ be the fraction of portfolio value invested in asset $i$, with ${\sum }_{i=1}^n{w}_i=1$, and let $R_i$ denote its return. For fixed weights over the measurement period, portfolio return is $R_P={\sum }_i{w}_i{R}_i$. If ${\mu }_i=E\left[R_i\right]$ and ${\sigma }_{ij}=Cov(R_i,R_j)$, then:

The vector $𝐰$ contains the weights and the matrix $\Sigma$ contains covariances. Its diagonal entries are individual variances. Off-diagonal entries satisfy ${\sigma }_{ij}={\rho }_{ij}{\sigma }_i{\sigma }_j$, where ${\rho }_{ij}$ is the return correlation. Portfolio risk is therefore not a simple weighted average of individual volatilities. Diversification depends on how assets move together, and a correlation estimated in calm conditions may not describe a stressed market.

The pairwise correlations are ${\rho }_{12}=0.50$, ${\rho }_{13}=0.20$, ${\rho }_{14}=0.30$, ${\rho }_{15}=0.50$, ${\rho }_{23}=0.40$, ${\rho }_{24}=0.30$, ${\rho }_{25}=0.10$, ${\rho }_{34}=0.70$, ${\rho }_{35}=0.15$ and ${\rho }_{45}=0.20$. Correlations are symmetric and each asset has correlation 1 with itself. Converting percentage returns to decimals before multiplying gives the following covariance matrix.

For example, ${\sigma }_{34}=0.70\left(0.18\right)\left(0.25\right)=0.0315$. Multiplying the expected returns by their weights gives ${\mu }_P=0.0500055$, approximately 5.00%. The quadratic form gives variance 0.0093482194 and standard deviation 0.0966862, approximately 9.7%. These numbers refer to the same horizon as the inputs. A spreadsheet implementation uses a weighted sum for the mean and matrix multiplication for ${𝐰}^{𝖳}\Sigma 𝐰$; the matrix should be symmetric and positive semidefinite. Rounding its entries too aggressively can distort the result.

Normal quantiles and a portfolio VaR calculation

If the vector of asset returns is jointly normal, every fixed linear combination is normal. It is not enough that each asset has an individually normal marginal distribution: the joint dependence structure also matters. Under the joint-normal assumption, $R_P\sim N({\mu }_P,{\sigma }_P^2)$. If $Z\sim N(0,1)$, then $R_P={\mu }_P+{\sigma }_{P}Z$. Approximately 5% of standard-normal observations lie below −1.645 and 5% above 1.645.

The lower 5% portfolio-return quantile is consequently ${\mu }_P-1.645{\sigma }_P$. For initial value $V_0$ and simple return over the chosen horizon, the corresponding normal loss quantile is:

Loss is measured relative to initial value here. A risk system that measures losses relative to expected value would report a different convention. The formula can also produce a negative loss quantile when expected gains dominate the selected threshold; reporting systems sometimes floor VaR at zero, but that convention should be stated rather than silently changing the mathematics.

Open figure at full size ↗

Suppose EUR 2 million is invested and the volatility over the risk horizon is taken as the rounded 9.7%. With expected return assumed zero for this separate VaR illustration, the 95% lower return quantile is $-1.645\left(0.097\right)=-0.159565$. The loss threshold is therefore $2,000,000\left(0.159565\right)=319,130$ euros, approximately 15.96% of value. This calculation deliberately uses rounded volatility and a zero mean; it does not simultaneously use the five-asset portfolio's 5% expected return.

The horizon cannot be inferred from the final amount. If 9.7% is annual volatility, the result is an annual-horizon calculation; if it describes another horizon, that horizon applies. Scaling volatility by the square root of time requires assumptions about independent increments and stable variance. Fat tails, changing volatility and nonlinear payoffs can make the normal approximation inadequate even when the arithmetic is correct.

Hedging systematic equity risk with index futures

An equity portfolio's systematic exposure can be measured by beta, ${\beta }_P=Cov(R_P,R_M)/Var\left(R_M\right)$, where $R_M$ is the market return. A beta of 0.8 means that a one-percentage-point market movement is associated, in the fitted relationship, with approximately a 0.8-percentage-point portfolio movement. It is a sensitivity, not a guarantee that every realised return follows the index. Firm-specific events and changing relationships remain possible.

A manager holding equities can reduce this exposure by selling appropriate stock-index futures. A short futures position gains when the index falls and loses when it rises. Index hedging can be quicker and less costly than selling every holding or hedging each security separately. The relevant index must resemble the portfolio's market exposure: using a DAX-related contract for German equity exposure and an S&P 500-related contract for US exposure are examples of the matching principle, not universal prescriptions.

For a contract with futures price $F_0$ and multiplier $m$, the notional exposure per contract is $m{F}_0$. A beta-based number of contracts to sell is:

where $V_P$ is current portfolio value and ${\beta }^{*}$ is the desired remaining beta. A positive $N$ denotes contracts sold. For market neutrality, ${\beta }^{*}=0$. Contract rounding, differences between the cash index and futures price, and basis changes prevent this formula from being a promise of exact protection.

Consider a USD 30 million portfolio with beta 0.8. Its beta-adjusted market exposure is USD 24 million. Using an illustrative E-mini S&P 500 futures level of 8,000 and a USD 50 multiplier gives USD 400,000 notional per contract, so the hedge sells 60 contracts. A 10% index fall implies an approximate USD 2.4 million systematic portfolio loss and an equal gain on the short futures. The combined systematic exposure is $0.8\left(30\right)-24=0$ million dollars.

Open figure at full size ↗

The USD 50 multiplier identifies the particular contract specification used in this example; other contracts have different sizes. A ten-point rise generates a USD 500 gain for the long side and an equal loss for the short side. Cash-settled futures do not require delivery of the underlying shares, and gains and losses are settled through daily variation margin. The familiar quarterly expiry cycle uses March, June, September and December contracts. Holding the hedge requires cash for adverse margin movements even when the combined investment performs as intended.

Market-neutral does not mean risk-free. The portfolio retains idiosyncratic risk, model and beta-estimation risk, basis risk and potentially liquidity risk. Beta can change as positions, prices and business conditions change. A manager must also decide whether to rebalance, roll an expiring contract or tolerate a temporary mismatch. The hedge is successful when it delivers the intended reduction in exposure at an acceptable total cost, not when its derivative leg always reports a profit.

Geometric Brownian motion and positive asset prices

Geometric Brownian motion, or GBM, is a tractable model for a positive asset price. It describes proportional changes rather than adding a normally distributed amount to the price regardless of its level. Its continuous-time specification is:

Here $S_t$ is price or portfolio value at time $t$, $\mu$ is the annualised drift, $\sigma$ is annualised volatility, and $W_t$ is a standard Brownian motion. An increment over $\Delta t$ years satisfies $\Delta W=\sqrt{\Delta t}\epsilon$, where $\epsilon \sim N(0,1)$. Thus the short-interval approximation is $\Delta S/S\approx \mu \Delta t+\sigma \sqrt{\Delta t}\epsilon$. The random term scales with the square root of time, not time itself.

The exact transition over a finite horizon $\tau =T-t$ is:

Equivalently, the log price is normal with mean $ln{S}_t+(\mu -{\sigma }^2/2)\tau$ and variance ${\sigma }^2\tau$. The price is lognormal, remains positive and is right-skewed. Finite-horizon simple returns are not exactly normal under this model; normality applies to log returns, while the short-interval proportional-change formula is an approximation. Using the exact exponential update avoids negative simulated prices that a coarse additive approximation could generate.

GBM is useful because its distribution and quantiles can be calculated directly and because it underlies a basic version of option-pricing theory. It remains an assumption. Constant volatility, continuous paths and independent Brownian increments omit volatility clustering, jumps, changing correlations and some features of extreme losses. For real-world risk simulation, drift represents the assumed physical return process. For risk-neutral derivative pricing, the drift is changed consistently with the pricing framework; expected investment return and the risk-free pricing drift should not be interchanged.

Monte Carlo simulation and the interpretation of a risk forecast

Monte Carlo simulation repeatedly draws random inputs and applies the model to generate many possible outcomes. Suppose a portfolio starts at EUR 10,000, with annual drift 9% and volatility 15%, and is projected over five years. A simulation of 10,000 paths can use monthly time steps, $\Delta t=1/12$, and the exact GBM transition. At each step, every path receives a fresh independent standard-normal draw. The result is a distribution of hypothetical model outcomes, not 10,000 forecasts that the market is expected to follow.

Digital investment services such as Scalable Capital have used distribution-based illustrations to explain possible investment outcomes to clients. The useful feature is the display of uncertainty around a central path rather than a single apparently certain future value. The quality of that communication depends on explaining the return and volatility assumptions and the difference between a model quantile and a promised investment result.

Analytical GBM quantiles provide a direct check on the simulation:

The median corresponds to $p=0.5$, whereas the expected value is $E\left[S_T\right]=S_0{e}^{\mu T}$. These differ because the lognormal distribution is skewed. A middle forecast labelled “50% quantile” is a median, not automatically the mean. The following comparison keeps analytical values separate from the estimates produced by one illustrative 10,000-path simulation.

The same illustrative simulation produced a five-year 95% quantile of EUR 25,644.96. Its difference from the analytical value is sampling variation, not a new economic assumption. More simulation paths generally reduce Monte Carlo noise, but do not correct an unsuitable model. A documented implementation should record the parameters, time convention and random seed, and compare simulated moments and quantiles with known results where available.

Open figure at full size ↗

A fan of 5%, 50% and 95% quantiles describes the distribution at each displayed date. It is not a 90% guarantee that an entire path stays inside the band at all intervening dates. Nor is the lowest simulated outcome the maximum possible loss. Distinguishing point-in-time quantiles, path-dependent losses and model uncertainty is essential when communicating risk forecasts to investment clients or decision-makers.

Risk-adjusted performance and incentives

High returns do not by themselves demonstrate skill. A portfolio can earn more because it takes more risk, employs more leverage or accepts an infrequent large loss. Risk-adjusted performance measurement compares reward with the relevant exposure and helps evaluate business units, managers and compensation. The chosen measure must match the decision: total volatility, systematic market risk and capital consumption answer different questions.

Under the capital asset pricing model's assumptions, efficient portfolios combining a risk-free asset and the market portfolio lie on the capital market line:

Here $R_f$ is the risk-free return over the same period, $E\left[R_M\right]$ and ${\sigma }_M$ are the market's expected return and volatility, and ${\sigma }_P$ is portfolio volatility. The Sharpe ratio measures excess return per unit of total volatility: ${SR}_P=(E\left[R_P\right]-R_f)/{\sigma }_P$. With illustrative market return 10%, risk-free return 3% and volatility 20%, it equals $(0.10-0.03)/0.20=0.35$. This is an example, not a permanent empirical Sharpe ratio guaranteed by investing in the market.

The security market line instead relates expected return to beta:

It applies in the CAPM to individual securities and portfolios, including inefficient portfolios, because diversifiable risk is not priced in that model. The Treynor ratio divides excess return by beta, ${TR}_P=(E\left[R_P\right]-R_f)/{\beta }_P$. Jensen's alpha measures return above the CAPM benchmark:

Sharpe is appropriate when the investor bears the portfolio's total risk; Treynor and alpha focus on systematic risk and depend on the benchmark model. Treynor becomes difficult to interpret when beta is near zero or negative. Ex-post estimated alpha is uncertain and can reflect an omitted risk factor rather than skill. Comparing ratios also requires consistent horizons, financing assumptions and return definitions.

For a bank business unit, risk-adjusted return on capital, or RAROC, commonly compares risk-adjusted income with allocated economic capital. The numerator should specify the treatment of expected losses, operating costs and taxes; the denominator should identify the capital convention. Comparing RAROC with a required return on capital links performance to shareholder value. Compensation based on this framework can be more informative than a bonus tied only to current profits, but it still requires independent risk measurement, limits, accountability and attention to losses that emerge after the measurement period.

Exchange-rate quotations and transaction exposure

Let $S_t$ denote units of home currency required to buy one unit of foreign currency at time $t$. For a euro-based company receiving dollars, $S_t$ is euros per US dollar. A rise in this rate means dollar appreciation against the euro; a fall means dollar depreciation. If a source instead quotes dollars per euro, the economic interpretation of an increase is reversed. The two rates are reciprocals, so their percentage changes are not exactly equal and opposite for large movements.

A fixed foreign-currency receipt $X$ due at time $T$ has home-currency value $X{S}_T$. Its sensitivity to the exchange rate is $X$, and a movement $\Delta S$ changes its value by $X\Delta S$. This relationship is exact when the amount received remains fixed. A foreign-currency payment has the opposite exposure from the payer's perspective: an appreciating foreign currency raises its home-currency cost. A forecast sale adds uncertainty about the amount and timing of the receipt, not merely the conversion rate.

Suppose a German manufacturer plans to sell 1,000 cars in the United States and expects USD 100 million from those sales. At EUR 0.90 per dollar, the expected conversion is EUR 90 million. If the dollar falls to EUR 0.82, the same dollar amount converts to EUR 82 million, a reduction of EUR 8 million. The dollar sales revenue has not changed; its euro value has. Conversely, a euro-based buyer owing USD 100 million benefits from the lower rate. Risk direction must therefore be established from the actual cash flow.

Open figure at full size ↗

Transaction exposure includes contracted receivables and payables as well as sufficiently identified planned transactions. The distinction matters for hedging. A derivative covering a fixed receivable offsets a known nominal amount. A derivative covering forecast sales can become an unbacked currency position if the sales fail to occur. Netting inflows and outflows in the same currency and period can reduce the amount requiring an external hedge, but netting should respect timing, legal entities and access to the relevant cash.

Operational choices before derivatives

Consider a stylised Porsche example in which the manufacturer produces in Germany, pays its operating costs in euros and sells cars both in the euro area and in the United States. A weaker dollar reduces the euro value of US receipts if dollar prices and quantities remain fixed. One response is to stop serving the US market, avoiding the exposure at the cost of abandoning revenue and strategic opportunities. Another is to raise dollar prices enough to preserve the euro revenue per car.

Passing currency changes to customers does not necessarily remove the business risk. Higher dollar prices can reduce sales when demand is price-sensitive or competitors do not raise their prices. A company may therefore replace exchange-rate risk per unit with quantity risk. Contracts that fix prices for long periods make adjustment slower; strong differentiation or market power can make it easier. Exposure depends on the interaction between currency, prices and volumes.

A natural hedge matches foreign-currency revenue with costs or financing in that currency. Production, procurement, sales activity or research spending in the United States can generate dollar costs against dollar receipts. Dollar borrowing creates dollar interest and principal payments that also offset some inflows. These decisions have consequences beyond currency risk: plant location affects productivity and supply chains, while borrowing affects leverage, maturity structure and credit risk. They should be evaluated as business choices rather than as costless substitutes for derivatives.

Financial hedges, including forwards, futures, options and currency swaps, can then address the residual exposure without moving production. A money-market hedge can create a similar economic offset through borrowing and lending in the relevant currencies. The appropriate combination depends on cash-flow certainty, horizon, market access, transaction costs and the firm's ability to fund collateral or settlement.

Transaction, translation and economic exposure

Transaction exposure focuses on the home-currency value of future receipts and payments. It is often measured over the next quarter or year because contracts and sales forecasts are more identifiable over shorter horizons. Translation exposure concerns the effect of exchange rates on financial statements when foreign-currency balances or the accounts of foreign operations are expressed in the reporting currency. Economic, or operating, exposure concerns the effect of currency movements on the firm's future competitive position and value.

The categories can overlap. A foreign subsidiary generates operating cash flows, holds balance-sheet positions and affects the group's competitive strategy. It is still useful to separate the objectives because a hedge designed to stabilise next year's euro earnings may not protect long-term firm value or a regulatory capital ratio. A policy that simply says “hedge currency risk” leaves the most important decision undefined.

Why hedging policies differ across firms

Historical survey evidence from M. Glaum's study of German non-financial corporations illustrates the diversity of practice. The percentages below describe the cited sample rather than current universal frequencies. Most respondents reported flexible hedging, while others used full coverage, a fixed proportion or no hedging. The distinction is economically meaningful: a rule-based hedge stabilises a defined exposure, whereas a flexible policy can incorporate changes in forecasts, costs or management views.

The horizon percentages sum to 101% because the reported figures are rounded. They should not be rescaled and presented as if more precise information were available. More fundamentally, a five-year hedge can be inappropriate when the quantity of sales five years ahead is highly uncertain. The problem is exposure estimation, even if a sufficiently long-dated derivative can be obtained.

Owners differ in diversification and risk preferences. A family whose wealth is concentrated in one firm may value stable cash flows differently from a globally diversified fund. Managers whose compensation depends on reported earnings may also dislike currency-related earnings surprises. Currency risk can be material for one business and negligible relative to other risks for another. Differences in demand elasticity, price-adjustment frequency, foreign costs and forecasting ability make identical hedge percentages unlikely to suit every firm.

When corporate hedging can create value

In an ideal frictionless capital market, changing the firm's currency payout pattern need not create shareholder value. Investors who can reproduce the hedge themselves at the same cost can choose their own exposure. A globally owned business also has no single currency that represents every shareholder's consumption or liabilities. Stable euro earnings may still be volatile in dollars for a US investor. Hedging is therefore not justified merely by saying that all volatility is undesirable.

Real-world frictions change the argument. Financial distress is costly, and a currency loss can cause a covenant breach, a rating downgrade or forced financing on unattractive terms. Stabilising operating cash flows can help preserve funding for positive-net-present-value investment when external finance is expensive. This is the investment-financing rationale associated with Froot, Scharfstein and Stein: the hedge protects the capacity to undertake valuable projects, not simply the appearance of smooth earnings.

Taxes can also matter when expected tax payments are a convex function of income, although the effect depends on the actual tax system, loss offsets and other circumstances. Firms may execute hedges more cheaply than small shareholders because of scale and market access. Managers and concentrated shareholders may be poorly diversified, creating both reasons to reduce distress and agency concerns about excessive hedging to protect their own positions. The value of a hedge must therefore be assessed against its costs and its effect on the firm's real decisions.

Forwards, options and the mechanics of an offset

A forward fixes today the rate $K$ at which currencies will be exchanged at a future date. A company expecting $X$ dollars can agree to sell them for euros at maturity. Combining the receipt with the short-dollar forward produces:

With USD 100 million and $K=0.90$ euros per dollar, the combined amount is EUR 90 million. If the spot rate at receipt is 0.82, the receivable converts to EUR 82 million and the forward contributes EUR 8 million. If the dollar strengthens to 0.98, the receivable converts to EUR 98 million and the forward loses EUR 8 million. The hedge removes the upside and downside around the contracted conversion rate. A derivative loss in the second scenario is the cost of the certainty previously chosen, not proof that the hedge failed.

A foreign-currency payable requires the opposite direction: a forward purchase of foreign currency. Its payoff is $X(S_T-K)$, which offsets the increase in the unhedged payment $X{S}_T$. A call option on the foreign currency offers the right to buy at a strike rate and therefore places a ceiling on the payment, while preserving the benefit of a cheaper foreign currency. The option buyer pays a premium for this asymmetry. A foreign-currency receivable can instead be protected with a put on that currency.

Futures are standardised and settled through daily margining; forwards are typically tailored over-the-counter agreements whose collateral terms depend on the contract. Both can create funding needs before the commercial exposure settles. Options have a different pattern of premiums and potential margin requirements depending on whether the company buys or sells them. A hedge policy should distinguish an option purchase that limits downside from writing options that may add exposures the business never intended to take.

Forward pricing, money-market hedges and cost of carry

A forward rate is not generally equal to today's spot rate and is not simply a prediction of the future spot rate. Under a simplified no-arbitrage framework with continuous compounding, home interest rate $r_h$, foreign interest rate $r_f$ and maturity $T$ years, covered interest parity gives:

The quotation remains home currency per foreign currency. The difference between interest rates determines the forward adjustment because holding foreign currency has a foreign-currency financing or investment return. The formula assumes consistent borrowing and lending opportunities and abstracts from transaction costs, credit effects and cross-currency basis. Those assumptions explain the economic benchmark; observed quotes may reflect additional market frictions.

A money-market hedge reproduces the same logic. To hedge a foreign-currency receipt $X$ at $T$, borrow its foreign-currency present value, convert that amount into home currency today and invest the home proceeds until $T$. The actual receipt repays the foreign loan. Using the stated continuous rates, the home proceeds are $X{e}^{-r_{f}T}{S}_0{e}^{r_{h}T}=X{F}_{0,T}$. The strategy uses borrowing capacity and changes the balance sheet, so an economic equivalence does not make every implementation equally convenient.

The related cost-of-carry principle applies to a non-income-paying investment asset. If it can be bought for $S_0$ and financed at continuously compounded rate $r$, a simple benchmark is $F=S_0{e}^{rT}$. For a spot value of EUR 80, an annual financing rate of 5% and delivery in six months, $F=80e^{0.05\left(0.5\right)}\approx 82.03$ euros. This example excludes asset income, storage costs and convenience benefits. It must not be confused with the currency formula, in which the foreign interest rate plays the role of a yield on the asset being held.

Translation exposure and the liquidity cost of hedging

A foreign operation may keep its accounts in a functional currency different from the group's presentation currency. Translating those accounts can change reported assets, liabilities and equity even when no sale or cash transfer occurs. If dollar-denominated net assets lose euro value as the dollar weakens, the group's reported equity can decline. The resulting effect on leverage or a covenant can matter even if the subsidiary's local operations are unchanged.

Under IAS 21, ordinary translation of a foreign operation into a different presentation currency generally places resulting exchange differences in other comprehensive income, or OCI, subject to the standard's conditions. This is distinct from remeasuring a foreign-currency monetary receivable in the entity's own functional currency, where exchange differences generally enter profit or loss. Treating every foreign-currency accounting change as the same type of translation exposure obscures this distinction.

Hedging a net investment can create a mismatch between realised cash and unrealised accounting gains. Suppose dollar appreciation raises the euro-translated value of a subsidiary's net assets by EUR 1 billion. A short-dollar hedge intended to offset that rise may require cash settlement or collateral even though the subsidiary has not distributed any cash. The accounting offset can be effective while the parent faces a liquidity demand. This is one reason some translation exposures remain unhedged.

It is too strong to say that translation exposure never affects earnings or is never hedged. Disposals, transaction balances, different functional-currency circumstances and designated hedging relationships can change the accounting treatment. More importantly, a firm may wish to protect a capital ratio or debt capacity rather than current profit. The decision should identify the relevant objective and the cash needed to support the hedge through adverse currency movements.

Why matching foreign assets and liabilities does not fix every ratio

Borrowing in the currency of foreign assets can reduce the sensitivity of net assets. If dollar assets and dollar liabilities are equal in amount, a change in their common euro conversion rate leaves their net contribution to equity approximately unchanged. Yet the gross balance-sheet amounts can still expand. With equity held constant, an increase in total assets raises assets-to-equity leverage and reduces an equity-to-assets capital ratio.

This distinction also applies to the regulatory leverage ratio. In that context the numerator is defined Tier 1 capital and the denominator is a regulatory exposure measure, not simply accounting assets. A stronger foreign currency can increase the denominator even when matched assets and liabilities offset economically. It is therefore essential to state whether “higher leverage” means a higher assets-to-capital multiple or a lower capital-to-exposure ratio.

A historical example is Deutsche Bank's first-quarter 2015 reporting: leverage exposure was EUR 1,549 billion at 31 March 2015, an increase of EUR 104 billion from year-end 2014, attributed almost entirely to exchange-rate movements. The figures illustrate how currency translation can affect a regulated denominator. They do not show that the bank incurred a EUR 104 billion loss, and they should not be interpreted as a current balance-sheet measure.

Foreign-currency instruments and accounting measurement

Accounting classification determines how price changes and exchange differences reach the statements, but banking-book and trading-book labels do not alone resolve every measurement question. Under IFRS, an instrument held for trading is normally measured at fair value through profit or loss. Its home-currency value can change because of the instrument's foreign-currency price and because of the exchange rate. Both components matter even when the issuer's contractual payment remains unchanged.

A foreign-currency government bond qualifying for amortised-cost treatment under IFRS 9 is different. A pure change in market yields does not by itself remeasure its carrying amount to the new market price. Effective-interest accounting, expected-credit-loss requirements and foreign-currency remeasurement still apply. As a monetary item, it is normally translated at the closing rate under IAS 21, with exchange differences generally recognised in profit or loss, subject to specified exceptions. “Held to maturity” is therefore not a reason to ignore exchange-rate movements.

German bank accounts under the HGB require a separate analysis. Section 340e(3) requires financial instruments in the trading portfolio to be measured at fair value less a risk deduction. This is different from a blanket rule that recognises all losses but never unrealised gains. For other assets, classification and the relevant valuation provisions matter. Currency translation must then be considered separately: section 256a provides a closing-spot-rate rule, while section 340h contains bank-specific provisions for positions specially covered in the same currency.

A general statement that a German bank's long-term foreign bond is unaffected by exchange-rate changes unless it suffers permanent impairment is consequently unreliable. Asset valuation, currency conversion, maturity and currency matching are separate parts of the analysis. The accounting result should also be separated from the economic exposure: a stable carrying amount does not mean the position could be sold at that amount or that its future cash flows are riskless in home currency.

Accounting references: IAS 21; HGB section 340e; HGB section 256a; HGB section 340h.

Economic exposure and operational flexibility

Economic exposure concerns changes in future cash flows and firm value caused by currency movements, including effects that are not attached to a booked receivable or payable. A European motorcycle producer can face yen exposure even if it sells few motorcycles in Japan and invoices almost nothing in yen. Japanese competitors pay important costs in yen. Yen depreciation can improve their ability to reduce prices abroad, putting pressure on the European producer's volumes or margins. Yen appreciation can reverse part of that competitive advantage.

The BMW motorcycle example illustrates this mechanism across changing competitive conditions: a largely domestic cost base does not eliminate exposure to the currencies of foreign competitors. A full historical account would also consider technology, product quality and market strategy, so exchange rates should not be presented as the sole explanation of decades of competitive performance. The essential distinction is between direct currency cash flows and indirect competitive sensitivity.

Operational hedging changes the structure of the business. Production and labour costs can be located closer to the markets generating revenue. Procurement can shift toward suppliers in the relevant currency, and shorter sourcing contracts may preserve flexibility. Maintaining the ability to move production between locations creates a real option: management can respond to future conditions instead of committing irrevocably to one cost structure. That flexibility has costs, including spare capacity and organisational complexity.

Financing can also be matched to durable operating inflows through foreign-currency debt or currency swaps. However, a financial derivative is usually tied to a fixed nominal amount and horizon, whereas economic exposure depends on changing prices, quantities and competitive responses. A large hedge based on a weak long-range sales forecast can increase rather than reduce risk. Scenario analysis should therefore examine the business response, not just mechanically translate a fixed cash-flow forecast at different rates.

Transfer restrictions and exchange-rate regime changes

Currency convertibility and the ability to transfer money are separate from the exchange rate itself. A subsidiary can earn profits abroad while capital controls, administrative restrictions or political decisions prevent distribution to the parent. Concerns reported about the transfer of profits from China illustrate this distinction. A derivative that fixes the conversion rate does not necessarily provide legal access to the underlying cash or permission to remit it.

Exchange-rate regimes can also change abruptly. The Swiss franc's history illustrates both safe-haven demand during financial stress and the risk of assuming that an official exchange-rate policy will persist. The Swiss National Bank's removal of its minimum exchange rate against the euro in January 2015 caused a sharp repricing. A period of low observed volatility under a policy regime may therefore understate the losses possible when that regime changes.

Open figure at full size ↗

These risks call for controls beyond a statistical estimate based on recent returns. Treasury should consider currency access, legal entities, settlement channels and the possibility that markets become discontinuous. Liquidity planning must allow for adverse collateral calls or the need to obtain currency when ordinary dealing conditions are unavailable.

Historical source: SNB, discontinuation of the minimum exchange rate, 15 January 2015.

The Lufthansa aircraft-purchase decision

The Lufthansa case concerns a decision made in January 1985 to purchase 20 Boeing 737 aircraft for USD 500 million, payable approximately one year later. The dollar had appreciated substantially against the Deutsche mark. The central question was how to manage the mark cost of a fixed dollar payment while preserving the airline's ability to finance its aircraft. A view that the dollar looked expensive was not a guarantee that it would decline before settlement.

For a transparent numerical comparison, let $S_T$ be the terminal exchange rate in Deutsche marks per dollar, assume a forward rate of DM 3.2 per dollar, and compare four strategies: no hedge, full forward cover, half forward cover and a call option on dollars. For the option example, assume a strike of DM 3.2 and a total premium of DM 100 million. The premium is an explicit illustrative input; its financing cost is ignored so that the payoff comparison remains clear.

With $X=500$ million dollars, the unhedged payment is $X{S}_T$. A full forward fixes it at $XK=1,600$ million marks. A half hedge fixes USD 250 million at 3.2 and buys the remaining USD 250 million at spot. A dollar call permits purchase at the lower of the spot rate and strike, but requires the premium. In DM billions, the four total costs are:

Open figure at full size ↗

The full forward removes exchange-rate uncertainty in this simplified calculation. The half hedge leaves half the slope of the unhedged cost curve. The option caps the total at DM 1.7 billion while retaining gains from a cheaper dollar, after allowing for its premium. The dollar protection is a call on dollars, equivalently involving a put on marks under the reciprocal currency description; calling it a dollar put would reverse the protection.

Ex-ante protection and ex-post regret

The dollar subsequently fell from approximately DM 3.2 to DM 2.2. Buying USD 500 million at the new spot rate would cost DM 1.1 billion instead of DM 1.6 billion. A half forward hedge at 3.2 generates a derivative loss of $250(3.2-2.2)=250$ million marks relative to spot, leaving a combined cost of DM 1.35 billion. The hedge gives up half the favourable currency movement in return for having protected half the payment against an unfavourable movement.

Open figure at full size ↗

Criticism based only on the realised forward loss ignores the decision's original purpose. Had the dollar risen to DM 4, the half hedge would have gained DM 200 million and reduced the DM 2 billion spot cost to DM 1.8 billion. The appropriate ex-ante assessment asks whether the chosen exposure, certainty and cost suited the airline's objectives using information available at the time. The strategy need not be optimal merely because it was a hedge, but hindsight about the direction of the dollar is not a sufficient performance criterion.

The case also highlights governance. Management and the supervisory board need agreement on whether the objective is to fix an investment budget, protect cash flow, stabilise accounting earnings or express a currency view. A partial hedge can be sensible under a defined policy, but looks arbitrary when no one can explain why that proportion was chosen. Clear reporting should distinguish the derivative result, the underlying purchase cost and the combined outcome.

Why the aircraft and the hedge can affect earnings differently

Buying an aircraft normally exchanges cash for a non-financial asset recognised at its acquisition cost. In a simplified initial-recognition example, that transaction does not immediately produce an operating profit or loss. A cheaper aircraft leaves more cash and a lower recorded asset cost; it does not necessarily create a current-period accounting gain equal to the currency saving. Depreciation and later measurement effects occur subsequently.

This balance sheet deliberately excludes derivative settlement and later depreciation. A 50% forward hedge of USD 500 million at 3.2 produces a DM 0.2 billion gain if the terminal rate is 4, and a DM 0.3 billion loss if it is 2. Without an accounting treatment that appropriately aligns the hedge with the exposure, the derivative result can enter earnings while the purchase-price difference is reflected in the aircraft's carrying amount. The economic offset and the timing of reported earnings can therefore differ.

Modern hedge accounting addresses specified mismatches when its designation, documentation and effectiveness requirements are met; it is not automatic simply because management calls a transaction a hedge. Nor should a historical case be used to claim that no form of hedge accounting existed anywhere in 1986. The durable lesson is to analyse both economic protection and accounting presentation before entering the transaction.

Putting the currency policy together

A corporate currency policy starts with the business decisions that generate risk: markets served, locations of operations and currencies used for pricing and financing. It then defines which transaction, translation and economic exposures matter, how confidently they can be forecast, the horizons to be covered and the permitted hedge percentages. The instrument choice follows these objectives rather than preceding them.

Limits can be expressed through nominal exposures, Value at Risk, Cash Flow at Risk or Earnings at Risk, depending on the quantity the firm seeks to protect. Operational hedges, foreign-currency borrowing, money-market transactions and derivatives should be considered together. Treasury also needs a process for exposure updates, collateral funding, counterparty limits and exceptions. A policy that reduces reported currency volatility while creating an unmanageable cash requirement has solved the wrong problem.

Transaction exposure is often the most straightforward to hedge because its amount and date can be identified. Translation exposure requires special attention to the difference between accounting gains and available cash. Economic exposure is often substantial but less directly matched by fixed-notional derivatives. The overall objective is a coherent relationship between the firm's operations, financial capacity and chosen currency risk, with the combined position understandable to those responsible for oversight.

Trading rewards, operating exposures and the purpose of a hedge

In December 2023, reporting on Goldman Sachs trader Ed Emerson put his three-year compensation near USD 100 million, illustrating potential commodity-trading rewards. Such examples must be distinguished from the objective of a corporate hedging programme. A trader may earn money by supplying liquidity, intermediating client transactions or deliberately accepting market exposure. An operating company normally hedges to stabilise an input cost, selling price, cash flow or balance-sheet value. The fact that a derivative loses money does not establish that the hedge failed: that loss may accompany an economically valuable gain in the underlying business.

A producer expecting to sell a known quantity is adversely exposed to falling prices; a consumer expecting to purchase it is adversely exposed to rising prices. The producer generally needs a short price hedge and the consumer a long price hedge. This directional rule applies to the specified exposure, not automatically to the company's name or industry. An oil distributor that has promised fixed-price deliveries but has not yet acquired the oil is exposed to rising acquisition costs, even though it will ultimately be selling the commodity to customers.

The risk-management objective should specify the quantity, delivery location, quality, currency and dates of the exposure, together with the loss measure and horizon. Hedging a price per barrel does not by itself hedge an uncertain number of barrels. Likewise, a crude-oil contract does not exactly hedge jet fuel, which includes refining and product-specific market effects. These distinctions explain why an apparently simple one-for-one hedge can leave substantial risk.

Forward contracts, futures and settlement obligations

A long forward commits its holder to buy a specified underlying at a delivery price agreed today; a short forward commits its holder to sell. Let $t$ be the agreement date, $T$ the delivery date, $S_T$ the spot price at delivery and $K$ the agreed delivery price per unit. Ignoring costs and default, the terminal economic payoffs are

For physical delivery, the long party pays $K$ and receives an asset worth $S_T$. Cash settlement pays the economically equivalent difference when the contract specifies it. These are linear payoffs: a one-unit increase in the terminal price raises the long payoff by one unit and lowers the short payoff by one unit. For $q$ units of the underlying, multiply the payoff by $q$.

A forward agreed at the current fair delivery price normally has zero initial market value. Zero initial value does not mean zero risk, zero future liability or unlimited access to funding. A contract agreed away from the fair price has a nonzero value, and collateral arrangements can require cash before maturity. Bilateral forwards also expose each party to the possibility that an in-the-money counterparty fails to perform.

Exchange-traded futures are standardised contracts supported by a clearing and margin system. Initial margin is a performance deposit; variation margin transfers gains and losses as the contract is marked to market. The clearing structure substantially reduces bilateral counterparty exposure, but it does not eliminate every clearing-member, operational, liquidity or extreme default risk. Futures may end in physical delivery or cash settlement. Closing a position before expiry requires an offsetting trade in the relevant contract, not merely a decision to stop recognising its value.

How a matched forward fixes a purchase cost

Suppose a consumer must buy one unit at $T$ and enters a long forward with $K=80$ currency units. The physical purchase costs $S_T$, while the forward produces $S_T-80$. The net cost after the hedge is therefore

If the price rises to 100, the physical cost is 100 and the derivative gain is 20. If the price falls to 60, the physical cost is 60 and the derivative loss is 20. Both cases give a net cost of 80. Evaluating only the derivative in the second case would miss the offsetting saving on the actual purchase. A producer selling one unit would instead combine sale revenue with a short forward, giving $S_T+(80-S_T)=80$.

Open figure at full size ↗

The cancellation requires the purchase and hedge to concern the same underlying, quantity and date and assumes performance by the counterparties. A similar example is a payment of USD 1 million due in three months hedged by a three-month forward that supplies exactly that dollar amount. The exchange-rate exposure of that payment can be removed, although funding, credit and operational risks remain. In a commodity business these matching conditions are often harder to achieve.

Cash-and-carry pricing and its assumptions

For a freely tradable asset that pays no income and has no storage costs, spot and forward prices are linked by financing. Let $S_t$ be the current spot price, $F_{t,T}$ the fair forward delivery price, $r$ a continuously compounded financing rate per year and $\tau =T-t$ the remaining time in years. Buying the asset today and financing it until delivery costs $S_t{e}^{r\tau }$ at $T$. Equivalently, a fair long forward plus an investment that grows to the delivery payment produces the same terminal asset. No arbitrage therefore requires

If a forward price were above this financing-adjusted spot price, an investor able to borrow, purchase and store the asset could sell it forward and lock in a difference. The reverse trade requires the ability to short or otherwise obtain the underlying. Transaction costs, credit limits and restrictions on short selling can therefore create bounds rather than an exact relationship in practice.

For example, with a spot price of EUR 80, a continuously compounded annual financing rate of 5% and six months to delivery, the no-income, no-storage calculation is $80e^{0.05\times 0.5}=82.0252$, or approximately EUR 82.03. The difference from the spot price is financing compensation, not automatically a forecast that the future spot price will rise by that amount. Forward prices and expected future spot prices need not coincide because risk premia and other economic forces can matter.

Valuing an existing forward before maturity

The delivery price of an existing contract remains fixed while the fair delivery price for new contracts changes. Suppose an earlier long forward has delivery price $K$. At a later date $s$, an opposite short forward with the same underlying and maturity can be entered at the new fair price $F_{s,T}$. At maturity, the physical obligations offset and the combined payoff is $F_{s,T}-K$. Under the simple deterministic-rate assumptions, the current value of the original long forward is

If the original delivery price was USD 10 and the new forward price is USD 11, the offset locks in USD 1 per unit at maturity. Its value today is the discounted dollar, not automatically exactly USD 1. The offset removes the price exposure but does not necessarily terminate both contracts or remove their counterparty and collateral obligations. Close-out and netting arrangements determine the contractual result.

Changes in forward prices are consequently a useful approximation to changes in forward value when discounting effects over the interval are small. Futures variation margin, by contrast, is paid or received during the contract's life. Under deterministic interest rates and otherwise identical terms, standard forward and futures prices coincide; with stochastic rates, the relationship between settlement cash flows and rates can produce a difference. The timing of gains and losses is economically relevant even when terminal price exposure appears similar.

Storage costs, convenience yield and the forward curve

Physical inventory can be costly to finance, insure and store, but possessing it can also provide an operating benefit. A refinery with available crude can keep producing during a supply disruption; a firm without inventory may miss a sale or interrupt production. Convenience yield represents this noncash benefit of holding the physical commodity. With proportional storage costs $u$ and convenience yield $y_c$, both expressed as annual continuous rates, a common cost-of-carry representation is

Financing and storage tend to raise the forward price relative to spot, while convenience yield tends to lower it. Contango commonly describes a forward price above spot or an upward-sloping curve across maturities; backwardation describes the opposite configuration. These related usages should be stated clearly. A high convenience yield can support backwardation despite positive financing and storage costs. As a deliverable contract reaches maturity, its price converges to the corresponding spot delivery value under the contract's delivery and settlement rules.

The formula is most informative when the underlying inventory and arbitrage assumptions are credible. Electricity is difficult to store economically at large scale, agricultural goods have seasonality and deterioration, and livestock involve biological production and maintenance. Location, quality and transport constraints further segment markets. Spot and deferred prices can consequently respond differently to the same news. That does not mean that the markets are literally unrelated; it means that a simple immediate storage trade may not enforce a tight price link.

For positive observed prices, an implied convenience yield can always be calculated as $y_c=r+u-ln\left(F_{t,T}/S_t\right)/\tau$. If this unknown is chosen solely to fit the observed forward price, however, the equation becomes a decomposition rather than an independent prediction. A collapse in current oil demand can depress nearby prices more than deferred prices as inventories accumulate and storage becomes scarce. The change in curve shape is then as important to a rolling hedge as the change in the overall price level.

Open figure at full size ↗

The February 2020 oil-curve comparison illustrates this distinction: nearby and deferred WTI prices both moved, but the size of the decline depended on the delivery month. Comparing entire curves reveals changes in term structure that a single spot-price observation would conceal. These are dated market observations, not a prediction of how every future demand shock must affect the curve.

Basis risk and the minimum-variance hedge

A cross-hedge uses an instrument that is related to, but not identical with, the exposure. Hedging jet fuel using crude-oil futures is an example: the jet-fuel bill and crude futures usually respond to common energy factors but also have separate influences. A hedge can also have maturity basis risk when short-dated contracts are rolled against a long-dated exposure. The relevant relationship is between changes over the actual risk horizon, not simply between long-run price levels.

Let $\Delta X$ be the unhedged change in the value of the business exposure, in currency, and let $\Delta H$ be the profit on one long futures contract over the same interval, also in currency. Let $N$ be the signed number of contracts: positive means long and negative means short. The hedged value change and its variance are

where ${\sigma }_X^2=Var\left(\Delta X\right)$, ${\sigma }_H^2=Var\left(\Delta H\right)$ and ${\sigma }_{XH}=Cov(\Delta X,\Delta H)$. Differentiating with respect to $N$ gives $2N{\sigma }_H^2+2{\sigma }_{XH}=0$. If the hedge contract has positive variance, the unique minimum is

where $\rho$ is the correlation of exposure changes and long-contract profits. The minus sign makes the position offset the exposure. A long inventory position generally has positive covariance with a long commodity future, requiring a short hedge. A future purchase cost enters business value with a negative sign; its adverse exposure generally calls for a long hedge. Defining the exposure as a cost instead of value is possible, but then the hedge payoff must be subtracted rather than added.

Contract size, regression estimation and remaining risk

If the exposure is $Q$ units of inventory at spot price $S$ and each futures contract covers $q$ units with futures price $F$, then $\Delta X=Q\Delta S$ and $\Delta H=q\Delta F$. The optimal signed number is

This distinguishes a dimensionless price hedge ratio from an actual contract count. For an illustrative inventory of 10,000 units, a contract size of 100 units, correlation 0.8 and a spot-change standard deviation 1.25 times the futures-change standard deviation, the price hedge ratio is $0.8\times 1.25=1$. The optimal position is therefore $N^{*}=-100$ contracts. Rounding, transaction costs and trading limits can make the feasible position differ from the continuous optimum.

Regressing $\Delta S$ on an intercept and $\Delta F$ estimates the slope $\beta =Cov(\Delta S,\Delta F)/Var\left(\Delta F\right)$. This is the positive hedge-size coefficient when covariance is positive; the signed inventory hedge is its negative after quantity conversion. The intercept captures an average change and does not itself reduce variance. Estimates should use observations relevant to the intended horizon and be checked for changes in market conditions rather than treated as permanent constants.

Substituting the optimal hedge into the variance expression gives

With a single regressor and an intercept, the regression's $R^2$ equals ${\rho }^2$ and measures the fraction of in-sample variance eliminated. Correlation 0.5 removes only 25% of variance; zero correlation offers no variance reduction. Perfect positive or negative correlation permits zero residual variance in this ideal linear model if either hedge direction is available. At correlation 0.7, 51% of the original variance remains and residual standard deviation is approximately 71.41% of its original level.

Open figure at full size ↗

If $R^2=0.50$, as in an illustrative single-stock hedge using index futures, half the variance remains but the remaining standard deviation is $\sqrt{0.50}=0.7071$, or about 70.7%, not 50%. This distinction is also essential when assessing commodity hedges. Variance efficiency measures a particular price-risk objective; it does not measure the ability to finance margin calls or guarantee that an estimated relationship will hold during a crisis.

When the underlying business disappears: Lufthansa in 2020

A forecast purchase is not the same as an unconditional obligation. The pandemic-related collapse in aviation activity reduced Lufthansa's fuel and currency needs, leaving some previously designated hedges without the forecast transactions they had been intended to offset. Falling fuel prices could then generate derivative losses without a matching saving on fuel that was no longer purchased. An excessive hedge became a separate open exposure.

Lufthansa's 2020 reporting identified EUR 758 million of realised fuel-hedge expenses and EUR 28 million of realised foreign-currency hedge earnings on expired stand-alone derivatives, a net result of EUR −730 million. These figures concern that specified category of derivatives, not the airline's total annual loss or its entire hedging result. The lesson is that volume forecasts, contractual commitments and hedge quantities must be reviewed together. A decision about whether to retain, reduce or close an overhedge also needs the resulting market-value and cash consequences.

Primary source: Lufthansa Annual Report 2020, page 213.

Metallgesellschaft: long-term promises and short-term cash demands

The Frankfurt-based industrial group Metallgesellschaft's oil-related crisis developed in 1993–94 and led to an emergency recapitalisation of the German parent company, rather than a simple completed bankruptcy. The core economic problem involved long-term fixed-price petroleum delivery contracts and large positions in short-term derivatives. The crisis combined price exposure, maturity mismatch, funding requirements and weaknesses in oversight; it cannot be understood by looking only at the final payoff of one futures contract.

A simplified delivery commitment makes the direction clear. If a distributor promises one barrel every month for ten years at a fixed selling price, it has committed to 120 future barrels. Higher purchase prices would reduce its delivery margin. Long oil futures can therefore offset the adverse price direction. But holding near-term futures for all 120 barrels concentrates exposure at the front of the curve, whereas the physical deliveries occur over 120 separate months. Short-term contracts must be rolled and adjusted as deliveries occur.

Open figure at full size ↗

When oil prices fell and the curve moved toward contango, long near-term futures generated losses that demanded immediate cash. Fixed-price customer sales could become more valuable because oil could eventually be purchased more cheaply, but those benefits were spread across future deliveries and were not equivalent to cash available for today's margin call. Different maturities also moved by different amounts, so an economically offsetting direction did not imply an equal offset in present value.

Rolling introduces another mismatch. A long position is closed in the expiring contract and reopened further out. In contango, the new contract may be more expensive than the old one, changing the economics of maintaining exposure. This difference is not necessarily an immediate payment of the whole futures notional: futures are margined contracts. Nevertheless, the evolution of prices, repeated settlements and financing of collateral determine the realised cash-flow path. A hedge that requires funding the company cannot obtain may be impossible to maintain.

Historical analysis: Frankel and Palmer, Federal Reserve International Finance Discussion Paper 560 (1996).

Evaluating hedge size and the complete risk budget

Historical regression estimates associated with the Metallgesellschaft debate illustrate why matching total barrels one-for-one need not minimise variance. The table relates changes in longer-dated futures prices to changes in the nearest one-month futures price over 1990–1992. The slope estimates below one indicate that a unit change in the near-term price was associated with a smaller change in the longer-dated contract; the relationship also varied by product and maturity.

For example, the six-month heating-oil slope of 0.513 implies a smaller offsetting near-term price exposure than a one-for-one position, under that regression's units and assumptions. Its $R^2$ of 0.80 implies 20% residual variance and about 44.72% residual standard deviation in the fitted single-factor model. These short-maturity historical estimates do not by themselves determine a ten-year delivery programme's optimal hedge. They demonstrate the need to estimate the relevant price relationship rather than assume equal sensitivities.

A hedge also needs clear responsibility for monitoring positions, approving exceptions and arranging collateral before a stressed market closes access to finance. A complete hedge policy combines exposure measurement, a defensible contract size, stress tests of curve changes and volume disruption, and a funding plan for adverse margin paths. Long-term delivery optionality and customer credit risk require separate analysis. Minimum variance is one useful objective, but management may instead need to protect minimum cash balances or avoid a severe funding shortfall. Those objectives can justify a hedge smaller than the unconstrained variance optimum, a different maturity structure or a different instrument.

Market-value risk and reinvestment risk

Market-value risk is the possibility that changing discount rates reduce the current value of an asset or increase the current value of a liability. For a conventional fixed-rate bond, higher required yields reduce present value because the same future payments are discounted more heavily. Reinvestment risk concerns the uncertain rates available when coupons, principal repayments or other cash receipts must be invested again. These risks can move in opposite directions: a fall in rates raises the current price of a fixed-rate bond but reduces the income available on future reinvestment.

Suppose an investor repeatedly purchases one-year instruments over a 30-year saving horizon. Ignoring costs and default, terminal wealth from an initial amount $W_0$ is $W_{30}=W_0{\prod }_{j=1}^{30}(1+r_j)$, where $r_j$ is the one-year rate available for year $j$. The first rate may be known, but the later rates are not. Short contractual maturity therefore does not eliminate uncertainty about long-horizon wealth. A pension institution must relate the timing and size of its assets to the timing and size of its benefit obligations.

The relevant objective might be tomorrow's liquidation value, next year's net interest income or the funding of a payment decades away. The same security can look relatively safe under one objective and risky under another. Inflation, credit risk and liquidity can add further uncertainty even when a contractual nominal payment is fixed. A duration calculation isolates a specified interest-rate sensitivity; it is not a complete statement about every risk of holding a bond.

Zero-coupon bond valuation and yield to maturity

A default-free zero-coupon bond makes one payment at maturity. Let $F$ denote its face value, $T$ its remaining maturity in years and $y$ its effective annual yield to maturity. With annual compounding, its price is

The yield is the discount rate that equates the contractual payment's present value to the observed price. It is quoted as an annual rate, while $y$ in the formula is a decimal: 4.68% is entered as 0.0468. For a payment of 100 in 20 years at exactly 4.68%, the calculation gives $100/{\left(1.0468\right)}^{20}=40.061363$, or 40.06 to two decimals. Conversely, a price of 40.04 corresponds to a yield of approximately 4.68279%, which also displays as 4.68% when the yield is rounded to two decimal percentage points. Retaining precision explains this otherwise apparent discrepancy.

The price decreases as the yield increases and the relationship is convex. At yields of 2%, 4% and 6%, the same 20-year payment of 100 is worth approximately 67.2971, 45.6387 and 31.1805. Longer maturities generally create greater percentage price sensitivity to the same yield movement, other things equal. A one-month zero-coupon payment is much less sensitive than a 20-year payment because discounting acts over a much shorter period.

Open figure at full size ↗

A single yield to maturity summarises the price of one instrument. It should not be mistaken for an assumption that every maturity in the market has the same discount rate. For more general fixed cash flows, valuation uses discount factors at each payment date; a yield to maturity is the single rate that reproduces the resulting total price under a chosen compounding convention.

Coupon bonds, par values and the term structure

A coupon bond distributes part of its value before the final redemption date. For annual coupons of amount $C$, face value $F$ and an integer maturity $T$, its single-yield pricing formula is

Each coupon and the principal must be discounted. When the coupon rate $C/F$ equals the required yield under the same payment convention, a conventional bond trades at par. A coupon rate below the required yield produces a price below par; a coupon rate above it produces a price above par. The discount or premium compensates the investor for the difference between contractual coupon income and the return required by the market.

Under a nonflat term structure, the more general representation is $P={\sum }_{j}C{F}_{j}d(0,t_j)$, where $C{F}_j$ is the cash flow at time $t_j$ and $d(0,t_j)$ is the corresponding discount factor. A five-year annual-coupon bond can be treated as a portfolio of five zero-coupon claims: the first four pay coupons and the fifth pays coupon plus principal. This decomposition becomes important when different parts of the yield curve move by different amounts.

Macaulay duration as present-value-weighted timing

Macaulay duration measures the average time at which a bond's value is received, using present values as weights. Let $PV\left(C{F}_j\right)$ be the present value of payment $j$, let $P={\sum }_{j}PV\left(C{F}_j\right)$ and define $w_j=PV\left(C{F}_j\right)/P$. Then

The weights sum to one for a positive-cash-flow bond, and duration is measured in years when payment times are measured in years. A ten-year zero-coupon bond has Macaulay duration ten years because all its value arrives at the final date. A coupon bond normally has a shorter Macaulay duration than its final maturity because some cash flows arrive earlier.

Portfolio duration is a market-value-weighted average under consistent conventions. Equal market-value investments in six-year and ten-year zero-coupon bonds give $0.5\times 6+0.5\times 10=8$ years of Macaulay duration. Equal face values would not generally give equal market-value weights. The distinction matters because risk sensitivity depends on today's economic values rather than simply the nominal amounts printed on the instruments.

The timing interpretation is particularly useful for relating assets to liabilities. Matching a portfolio's duration to the timing sensitivity of a future obligation can reduce first-order exposure to a small parallel rate change. It does not ensure that every cash flow occurs on the right date, eliminate reinvestment uncertainty under every rate path or provide protection against arbitrary changes in the curve's shape.

Modified duration and the value of a basis point

Modified duration converts cash-flow timing into first-order price sensitivity. For a nominal annual yield compounded $m$ times per year,

Here $m$ is the number of coupon or compounding periods per year, $y$ is expressed as a decimal annual rate, and $\Delta y$ is an absolute change in that rate. A rise from 4% to 5% is $\Delta y=0.01$, not 1 and not a 25% relative yield change. The negative sign means that a positive-duration asset loses value when its yield rises.

One basis point, abbreviated bp, is 0.01 percentage points, or 0.0001 in decimal rate units. The positive magnitude of the approximate loss from a one-basis-point yield increase is often called PVBP, DV01 or the value of a basis point:

For the 20-year zero-coupon example at exactly 4.68%, $D_{\mathrm{m}\mathrm{o}\mathrm{d}}=20/1.0468=19.105846$ and $\mathrm{P}\mathrm{V}\mathrm{B}\mathrm{P}=40.061363\times 19.105846\times 0.0001=0.076541$ currency units per bp. A ten-basis-point rise therefore implies an approximate loss of 0.7654. Some systems report a signed PV01 sensitivity rather than a positive loss magnitude; the convention must be checked before combining positions or calculating hedge quantities.

Convexity and the limits of a linear approximation

Duration is the tangent slope of the price–yield relationship at the current yield. It becomes less accurate as the rate change grows because the slope itself changes. Price convexity, denoted $𝒞$, is the second derivative of price divided by price. A second-order approximation is

For a zero-coupon bond under annual compounding, $𝒞=T(T+1)/{(1+y)}^2$. The 20-year example has convexity approximately 383.2850. For conventional positive fixed cash flows, convexity is positive: compared with the tangent approximation, the exact price is higher after either an upward or a downward yield shock. Duration alone therefore overstates the loss from a rise and understates the gain from a fall, for the same absolute shock.

Open figure at full size ↗

The second-order calculation is considerably closer here but remains an approximation. Instruments with options can have different curvature because their expected cash flows change when rates change. Moreover, one duration and one convexity measure describe movement along one specified yield direction. They do not automatically capture twists, steepening or changes in credit spreads.

Hedging a bond portfolio with futures

Consider a corporate-bond portfolio worth USD 10 million with modified duration nine years. A one-basis-point parallel increase in its required yields produces an approximate value change of $-10,000,000\times 9\times 0.0001=-9,000$ dollars. Suppose one long Treasury-bond futures contract loses approximately USD 90 for the corresponding one-basis-point increase. An offsetting hedge requires

The minus sign means 100 short futures contracts. If the matched rate movement occurs, the short futures gain approximately USD 9,000 while the bond portfolio loses approximately USD 9,000. This neutralises the specified first-order parallel-rate sensitivity. It does not make the entire portfolio risk-free or ensure zero profit and loss under all market changes.

Corporate-bond spreads can change relative to government yields; the portfolio and futures can have different convexity and maturity exposures; and the yield curve may move non-parallelly. The futures sensitivity can itself depend on delivery terms and the bond expected to be cheapest to deliver. Hedge ratios therefore require monitoring and rebalancing. Futures settlement also creates cash flows before bonds are sold, so sufficient liquidity is needed even when economic value changes offset.

Defining bond Value at Risk and the modelling horizon

Value at Risk, or VaR, is a selected quantile of a defined loss distribution over a defined horizon. If $L=-\Delta P$ denotes a positive loss and $\alpha$ is the confidence level, then ${\mathrm{V}\mathrm{a}\mathrm{R}}_{\alpha }=q_{\alpha }\left(L\right)$. A 95% VaR is not a worst possible loss: the model assigns 5% of outcomes to losses exceeding that threshold. The cash flows, valuation date, holding period and treatment of carry must be stated.

Four simple methods can be illustrated with the same zero-coupon pricing function. One can model yield changes or directly model repriced bond changes, and one can use empirical quantiles or a normal-distribution approximation. These choices affect the result independently. A nonlinear pricing function can transform a symmetric yield distribution into an asymmetric price-change distribution.

The numerical illustration uses weekly changes and therefore a one-week horizon. Mixing weekly volatility with a one-day VaR label changes the meaning of the result. The simple shock calculations below hold remaining maturity fixed while applying a yield change, isolating market-rate risk. A full one-week holding-period valuation should also reduce remaining maturity and include any cash flows or financing over the week; those adjustments must be applied consistently across all methods.

Historical yield shocks and the duration-based loss calculation

Suppose the empirical 95th percentile of weekly yield changes is an increase of 9.77 bp and the 5th percentile is −8.60 bp. For a long positive-duration bond, the upper yield tail is the adverse direction. Using the calculated PVBP gives

The equivalent duration calculation uses $\Delta y=9.77/10,000=0.000977$. The approximate fractional price loss is $19.105846\times 0.000977=0.0186664$, or about 1.8666%. Multiplying by the price 40.061363 gives the same loss of approximately 0.748. Entering 0.00977 would mean a 97.7 bp shock and make the first-order loss ten times too large.

This estimate combines an empirical yield distribution with a linear value approximation. It is conditional on the selected historical period, data quality and quantile convention. A change in those inputs can change the result without any change in the bond itself. The relevant adverse tail would reverse for a short position, and a portfolio with options might not have a globally monotonic response to the same yield factor.

Normal yield shocks and full historical repricing

For a zero-mean normal model of weekly yield changes with standard deviation ${\sigma }_{\Delta y,\mathrm{b}\mathrm{p}}$, the 95% first-order VaR is

With a weekly standard deviation of 5.443 bp, the result is $1.6448536\times 5.443\times 0.076540625=0.685263$ currency units. The factor 1.645 is a one-sided 95% normal quantile; using 1.96 would instead correspond to a different tail probability. If a nonzero expected change is included, it must be carried through the loss model rather than silently omitted.

Historical full repricing avoids the duration approximation. For each observed or specified shock $\Delta y_j$, calculate $P_j=100/{(1+0.0468+\Delta y_j)}^{20}$, then form the price change $\Delta P_j=P_j-P_0$ and loss $L_j=-\Delta P_j$. The following scenarios use yield changes expressed in bp; they are valuation calculations, not a claim that these few rows constitute the whole estimation sample.

For the same adverse 9.77 bp shock used in the linear model, full repricing gives $40.061363-100/{\left(1.047777\right)}^{20}=0.740523$. Positive convexity explains why this loss is slightly smaller than 0.747802. With a monotonic fixed-maturity price function, a consistently defined empirical upper yield quantile maps to the corresponding upper loss quantile, subject to the interpolation convention. One should not interpret a materially different value as a convexity effect without checking that the samples and signs match.

A normal model for price changes and numerical consistency

The fourth method estimates the standard deviation of repriced bond changes and then imposes a normal approximation directly on those changes. If their standard deviation is 0.417 currency units and their mean is neglected, the 95% VaR is $1.6448536\times 0.417=0.685904$. Yield-normal and price-normal models can give similar numbers for small shocks, but they are different assumptions because the bond price is a nonlinear function of yield.

An illustrative comparison based on weekly 2024 observations reports rounded VaR figures of 0.75 from the empirical yield quantile, 0.68 from the normal yield model, 0.66 from an empirical price-loss quantile and 0.69 from the normal price-change model. Its price-change summaries are a 5th percentile of −0.66, a 95th percentile of +0.72 and a standard deviation of 0.417. These are separate reported summary statistics, not enough information to reconstruct a complete common sample. In particular, the 0.66 empirical loss cannot be reproduced by applying the stated 9.77 bp adverse quantile to this fixed-maturity bond; the consistent calculation gives approximately 0.741.

The useful lesson is the method comparison, together with the requirement to reconcile inputs. A 95% loss quantile is the negative of the 5% price-change quantile, not the positive 95% price-change quantile. Small changes around 0.685 can also round to either 0.68 or 0.69 when underlying inputs have already been rounded. A robust calculation keeps unrounded prices and sensitivities, aligns observation frequency and sample dates, and applies a common loss convention before comparing model outputs.

Multiple yield-curve factors and instrument optionality

A coupon bond does not have only one economically relevant interest rate. Its component payments depend on discount factors at different maturities. Historical simulation can apply a joint observed change in the one-, two-, three-, four- and five-year rates to a five-year bond, then revalue all payments. The rate changes must come from the same historical observation so that their dependence is preserved.

For a first-order parametric model, let $\Delta 𝐲$ be a vector of decimal changes in the chosen curve rates, let $𝐠$ contain the corresponding currency sensitivities $\partial P/\partial y_k$, and let $\Sigma$ be the covariance matrix of those changes. Then

Adding separate maturity risks without covariance terms would miss the way curve movements reinforce or offset one another. Key-rate sensitivities make this structure visible. Similar valuation and scenario methods apply to interest-rate swaps and futures, with contract-specific cash flows and settlement conventions. Callable bonds, convertible bonds and bond options require additional modelling because exercise decisions or other risk factors can change the cash flows themselves.

Risk horizons, pull to par and a shrinking duration

A default-free zero-coupon bond pays its face value at maturity regardless of the path of interim market yields. As maturity approaches, its remaining duration falls and its price converges to that payment. Conditional on a similar distribution of yield shocks, its near-term interest-rate price sensitivity therefore decreases. This is a changing exposure, not a reason to apply the same volatility estimate mechanically at every future date.

If the horizon is exactly the maturity date, the nominal redemption cash flow is certain under the default-free assumption. The dispersion of that terminal payment due solely to interest-rate uncertainty is zero. It does not follow that every definition of investment loss or every investor's risk is zero: the comparison benchmark, purchase price, inflation and liquidity needs still matter. A coupon bond additionally creates reinvestment uncertainty for payments received before the horizon.

A short-term trader may need to know tomorrow's liquidation loss, whereas a saver concerned with retirement wealth in 2065 needs a model of cash flows and reinvestment over a much longer period. The horizon should follow the economic decision. Extending a short-horizon VaR by a square-root rule without considering declining maturity, changing sensitivities and cash flows can produce misleading results for bonds held over long periods.

Economic value of equity and the banking-book duration gap

Interest rate risk in the banking book, abbreviated IRRBB, concerns the effect of interest-rate movements on a bank's economic value and earnings outside its trading activities. Let $A$ and $L$ be the present values of assets and liabilities, and let $E=A-L$ be economic equity. If $D_A$ and $D_L$ are their modified durations under a common small parallel yield shock, then

The liability term must be subtracted because equity equals assets minus liabilities. A decline in the economic value of a liability benefits equity, other things equal, whereas a decline in asset value reduces it. Writing $D_{\mathrm{g}\mathrm{a}\mathrm{p}}=D_A-\left(L/A\right)D_L$ gives the equivalent expression $\Delta E\approx -A{D}_{\mathrm{g}\mathrm{a}\mathrm{p}}\Delta y$.

For an illustrative bank with assets of 100, liabilities of 90 and equity of 10, take asset duration five years and liability duration zero. A parallel rise of 200 bp, or 0.02, gives $\Delta E\approx -100\times 5\times 0.02=-10$, equal to the initial equity in magnitude. This is a first-order economic-value stress, not a statement that an accounting loss of exactly 10 must immediately be booked. If liability duration were two years instead, the approximate change would be $-(100\times 5-90\times 2)\times 0.02=-6.4$.

Basel's IRRBB framework considers both economic-value and earnings measures and includes parallel and non-parallel interest-rate scenarios. A single 200 bp shock is useful for exposition and historical comparison, but it is not a complete modern stress framework. Steepening, flattening and movements concentrated at short maturities can reveal exposures concealed by a small aggregate duration gap.

Regulatory reference: Basel Framework SRP98, guidance on interest-rate risk in the banking book.

Deposit behaviour and historical evidence from German banks

The effective sensitivity of retail deposits is difficult to infer from legal maturity alone. A sight deposit can be withdrawn immediately, but a stable portion of aggregate balances may remain with the bank for years. At the same time, the rate paid to depositors may adjust slowly or rapidly when market rates change. Balance stability, contractual withdrawal rights and the repricing of deposit rates are related but distinct modelling questions. An average historical holding period is not automatically the correct duration.

Assuming a long behavioural duration for deposits can reduce the estimated duration gap against long fixed-rate assets. If customers withdraw funds, demand higher rates or move from sight deposits to time deposits during stress, that assumed protection can weaken. Behavioural assumptions must therefore be tested across scenarios and linked to the bank's funding strategy rather than accepted merely because past balances appeared stable.

Open figure at full size ↗

Open figure at full size ↗

The historical German mortgage evidence shows a growing share of new household house-purchase lending with initial rate fixation longer than ten years during 2003–2019. The accompanying banking-sector measures show that duration exposure differed substantially across credit cooperatives, savings banks and commercial-bank groups. The composition of new loans is not identical to the duration of the whole existing balance sheet, but it helps explain how prolonged low-rate conditions could increase fixed-rate asset exposure.

Open figure at full size ↗

The Bundesbank's 2022 historical comparison expresses the present-value loss from a 200 bp upward shock as a percentage of own funds. Savings banks and credit cooperatives had larger reported ratios than the group of large systemically important banks in that series. These are category-level measures with specified weighting and reporting conventions, not the same as the loss rate on assets or a conclusion about every institution in each group.

Open figure at full size ↗

A separate March 2022 comparison for German banks under the Single Supervisory Mechanism contrasts reported deposit assumptions with fixed average maturities and legal maturity of zero for relevant nonbank euro deposits. The upper percentiles of measured risk rise substantially under the legal-maturity treatment. The comparison demonstrates model sensitivity: the economic-value estimate depends not only on observed market prices but also on assumptions about how deposit balances behave.

Repricing gaps and net interest income

Economic value is a present-value measure over future cash flows; net interest income, or NII, is interest revenue minus interest expense over a period. A bank can experience a fall in the market value of long fixed-rate assets while its current interest margin initially improves because some asset rates reset faster than deposit costs. Conversely, funding costs can increase before fixed asset income adjusts. The value and earnings perspectives are complementary, not interchangeable.

In a simplified one-period repricing model, let $RSA$ denote assets whose rates reset within the horizon and $RSL$ the corresponding liabilities. If both reprice immediately and fully to the same rate change, the approximate annualised income effect is

For hypothetical rate-sensitive assets of EUR 60 million and liabilities of EUR 80 million, the gap is EUR −20 million. A two-percentage-point increase implies an annualised NII change of $-20\times 0.02=-0.4$ million euros. Actual within-year effects depend on reset dates, rate pass-through, floors and business volumes. This gap is measured in currency and differs from the duration gap used to approximate economic-value changes.

Open figure at full size ↗

Open figure at full size ↗

The Bundesbank's comparison through the second quarter of 2024 distinguishes realised annual changes in net interest income from scenario calculations for 2024. It shows stronger improvements for savings banks and credit cooperatives after the interest-rate reversal, while the projected effect depends on customers moving funds from sight deposits into more highly remunerated time deposits. Scenario bars should not be read as subsequently realised outcomes.

Open figure at full size ↗

The longer 2014–2022 series shows declining NII relative to total assets for savings banks and credit cooperatives during much of the low-rate period. Large systemically important banks are shown on a separate scale and have different business mixes. The median, asset-weighted mean and interquartile range summarise different features of the distribution; comparing line heights without checking axes or weighting can give a misleading conclusion.

Amortised cost, hidden reserves and unrealised losses

Accounting carrying amounts and market values answer different questions. An instrument carried at amortised cost does not generally have its carrying amount reset to current fair value after every market-rate movement. Its economic value can nevertheless change materially. A market value above the relevant carrying value creates an unrealised gain, often described as a hidden reserve in this context; a market value below it creates an unrealised loss. Hidden reserves are not another name for losses.

Open figure at full size ↗

The historical Bundesbank series for German savings banks and credit cooperatives, expressed relative to Common Equity Tier 1 capital and extending to the second quarter of 2024, shows earlier positive differences turning negative around the 2022 rate increase. Its definition compares the banking book's present value with its book-value approximation. The chart illustrates a valuation gap under that stated methodology, not an automatic deduction of the whole amount from reported regulatory capital.

If a bank can hold assets and finance them as planned, an unrealised market loss need not require immediate liquidation. If deposits leave or collateral needs rise, sales can crystallise losses and reveal the connection between interest-rate risk and funding liquidity. Effective management therefore combines economic-value stress tests, earnings projections, behavioural deposit analysis and cash-flow capacity. The label “banking book” or “amortised cost” does not remove the underlying economic sensitivity.

Market value, terminal payoff and profit

An option creates a contingent right rather than an unconditional obligation for its buyer. A call gives its holder the right to buy the underlying at the strike price; a put gives the right to sell. The writer receives the premium and accepts the corresponding obligation. Let $S_t$ denote the underlying price at time $t$, $K$ the strike, $T$ the expiry date, and $C_t$ or $P_t$ the market value of a call or put. All prices and premiums in a particular example are expressed in the same currency per unit of underlying. Contract multipliers convert these unit amounts into the cash exposure of actual contracts.

The terminal payoff is the contractual amount received at expiry. Profit also accounts for the premium and, in a complete investment calculation, financing, transaction costs and any other cash flows. Unless stated otherwise, the terminal-profit examples below subtract the initial premium without accumulating interest. This convention makes the payoff geometry transparent; it is not a claim that financing is free in actual trading.

A call with strike EUR 25 purchased for EUR 2.50 can lose market value before expiry even while the underlying remains above EUR 25. Its resale price depends on the underlying, remaining time, implied volatility and interest rates. If the call expires with $S_T\le 25$, its payoff is zero and the buyer loses the EUR 2.50 premium. If $S_T=26$, it pays EUR 1 but the buyer still has a net loss of EUR 1.50. Ignoring financing, break-even is EUR 27.50. An option finishing in the money is therefore not necessarily a profitable investment.

Intrinsic value is the immediate-exercise value, $max(S-K,0)$ for a call and $max(K-S,0)$ for a put. The difference between market value and intrinsic value is conventionally called time value. For options with an immediate-exercise right it is non-negative under usual no-arbitrage conditions; for European options, discounting and exercise restrictions mean that comparison with immediate intrinsic value requires care. A positive remaining maturity should never be treated as guaranteeing a particular time-value sign for every contract.

Here $c$ and $p$ are premiums paid by buyers and received by writers; the put bounds assume a non-negative underlying price. For an underlying that can become negative, those bounds must be reconsidered. “Naked” describes an option position without the relevant offsetting exposure. A short call backed by shares has a different aggregate risk profile from a naked short call, even though the option leg itself is identical.

Open figure at full size ↗

A bull call spread: constructing a bounded position

A bull call spread expresses a moderately positive view while limiting both the premium at risk and the benefit from a large price increase. Buy a call with lower strike $K_1$ and sell a call with higher strike $K_2$, using the same underlying, expiry and unit quantity. If the premiums are $c_1$ and $c_2$, the net premium paid is $d=c_1-c_2$. The terminal profit is:

Consider a stock initially trading at USD 50. A strike-50 call costs USD 3, and a strike-55 call can be sold for USD 1. The initial debit is USD 2. Below the lower strike both calls expire worthless; between the strikes only the purchased call pays; above the higher strike the two exercise values rise at exactly the same rate, so their difference stops increasing.

The maximum loss is USD 2, the maximum profit is $55-50-2=3$ dollars, and break-even is $50+2=52$ dollars. The maximum payoff is USD 5, which differs from the maximum profit. These amounts are per unit: a spread using contracts covering 100 shares has a USD 200 initial debit and USD 300 maximum terminal profit, before fees and financing.

Open figure at full size ↗

The terminal bound does not describe every operational exposure. If an American-style short call is assigned early, the investor may acquire a stock delivery obligation before the long call is exercised or sold. Funding, dividends and transaction costs can then affect realised results. The economic spread should be analysed together, while exercise and settlement obligations must also be managed leg by leg.

Equivalent shapes and the bull put spread

A bull put spread buys the lower-strike put and sells the higher-strike put. Its terminal option payoff is $max(K_1-S_T,0)-max(K_2-S_T,0)$. This is the bull call spread’s payoff minus the constant $K_2-K_1$. Consequently, the two option packages do not have identical raw cash payoffs. Their risk-dependent shapes are the same, with the difference accounted for by a fixed payment and the corresponding financing.

For European options with common maturity, put–call parity links their initial prices. With a constant continuously compounded rate $r$, the call-spread debit equals the put-spread debit plus $(K_2-K_1)e^{-rT}$, taking the valuation date as zero. A matching risk-free cash position therefore converts one strategy into the other. Comparing only the names “calls” and “puts” conceals this equivalence; comparing only terminal profits without a financing convention can also conceal a genuine cash-flow difference.

At zero interest, the strike-50/55 call spread costing 2 corresponds to a bull put spread receiving a credit of 3. Below 50 the put spread pays −5, giving profit −2 after its initial credit; above 55 it pays zero and retains profit 3. The same break-even and maximum profit or loss follow. The example demonstrates why portfolio risk analysis must aggregate contractual cash flows and premiums before deciding whether a position is bullish, bearish or bounded.

Straddles, butterflies and the price of convexity

A long straddle purchases a call and a put with the same strike and expiry. Its terminal payoff is $|S_T-K|$, and its profit is $|S_T-K|-(c+p)$. The buyer benefits from a sufficiently large move in either direction, not merely from any movement. With total premium $a=c+p$, break-even prices are $K-a$ and $K+a$, provided the lower one is feasible for the underlying. The maximum loss is the premium, occurring when the terminal price equals the strike.

As an explicitly illustrative example, take $K=50$ and total premium 12. The break-even prices are 38 and 62. Expiry at 40 still produces a loss of 2 because the payoff of 10 does not recover the premium. Expiry at 70 produces profit 8. Before expiry, the value can increase when implied volatility rises even if the underlying price barely moves; an eventual terminal profit also depends on how much movement was already priced into the options.

A short straddle reverses these profits. The writer retains the total premium when the underlying finishes at the strike, but faces losses for sufficiently large moves. For a non-negative equity price the downside-price loss is bounded by $K-a$, while the upside-price loss is unbounded. A calm historical sample can make this strategy appear attractive even though a rare large move creates severe losses. Margin calls can force liquidation before the terminal payoff is reached.

A long call butterfly buys one call at $K_1$, sells two at $K_2$, and buys one at $K_3$, with equally spaced strikes and common expiry. Its payoff is:

The payoff is zero outside the two outer strikes and reaches $K_2-K_1$ at the middle strike. Subtract the net debit to obtain profit. With illustrative strikes 45, 50 and 55 and debit 2, the maximum loss is 2, maximum profit 3 and break-even prices 47 and 53. The purchased wings limit tail losses relative to an unprotected short-volatility position. Adding appropriate wings to a short straddle produces an iron-butterfly structure; a call butterfly is a related bounded structure, not literally a long straddle with two arbitrary extra options.

Open figure at full size ↗

The Greeks as local sensitivities

Terminal diagrams hold the expiry date fixed and vary the final underlying price. Daily risk management asks a different question: how much will the current option value change when market inputs move? Write $V=V(S,\sigma ,r,t)$, where $\sigma$ is annualised implied volatility, $r$ the relevant rate, and $t$ calendar time. A local Taylor expansion separates major sources of profit and loss:

The symbol $\Delta$ is used both for a Greek and, when preceding a variable, for a change; the product $\Delta \Delta S$ means option delta multiplied by the price change. If instead time to maturity $\tau =T-t$ is the pricing input, calendar-time theta is $-\partial V/\partial \tau$. Confusing these conventions reverses the reported sign. A theta calculated per year must also be converted consistently before applying a one-day change.

Volatility of 20% is 0.20 in decimal form. An increase to 21% is one percentage point, meaning $\Delta \sigma =0.01$, not 1. A raw vega derivative of EUR 500 per unit decimal volatility implies approximately EUR 5 for that change. Trading systems often report vega directly per percentage point and rho per basis point or percentage point. Units must therefore accompany each number. FX options may require sensitivities to both domestic and foreign interest rates. If $r^{*}$ is the foreign rate, the expansion adds ${\rho }^{*}\Delta r^{*}$ with ${\rho }^{*}=\partial V/\partial r^{*}$. Complex portfolios need term-structure and volatility-surface buckets rather than one universal rate or volatility.

The expansion omits higher powers and cross-derivatives, including interactions between price and volatility. Its accuracy deteriorates for large shocks, near discontinuities and when important inputs are missing. Gamma improves on delta alone, but delta–gamma is still an approximation. Full revaluation under joint scenarios is the appropriate comparison when assessing whether the local approximation remains useful.

Aggregating Greeks into a usable risk report

Portfolio Greeks add across positions when they refer to the same risk factor, units and valuation conventions. For signed quantities $n_j$ and contract multipliers $m_j$, portfolio delta is ${\sum }_j{n}_j{m}_j{\Delta }_j$, and analogous expressions apply to the other Greeks. Long and short positions can offset in one sensitivity while reinforcing another. A near-zero aggregate delta is compatible with substantial gamma or vega exposure.

Consider an explicitly illustrative portfolio worth EUR 150 million with 120 positions in equities, indices and commodities. To make an aggregate scenario report interpretable, define a common parallel relative-price shock $x=\left(\Delta S/S\right)/0.01$, a volatility shock $v=\Delta \sigma /0.01$, a rate shock $q=\Delta r/0.01$, and elapsed time $d$ in days. These are standardised scenario coordinates, not claims that different underlyings always move together. The following local coefficients are in EUR millions:

A 5% parallel price decline has $x=-5$: the delta contribution is −50 million and the gamma contribution is +25 million, giving a local delta–gamma estimate of −25 million. Reporting only −50 million is a first-order calculation, not a calculation that incorporates the stated gamma. A ten-percentage-point volatility increase gives a vega-only gain of 50 million, whereas a 10% relative increase from 20% to 22% volatility gives only 10 million. A half-percentage-point rate increase gives 1.5 million. These isolated calculations hold other inputs fixed; large joint stresses require full revaluation.

A practical report also separates underlyings, maturities and volatility buckets, identifies large concentrations, and shows the effect of proposed hedges. The number of positions and market value help describe scale but do not measure risk by themselves. Reporting a Greek simply as a currency amount without specifying its shock normalisation can produce errors larger than the apparent precision of the calculation.

Delta hedging and the residual gamma exposure

Suppose a purchased call has delta 0.6 and covers one share. A USD 1 increase in the share price gives an approximately USD 0.60 call gain. Shorting 0.6 shares creates an approximately USD 0.60 loss on the stock leg, cancelling the first-order effect. The hedge reverses for a short call: the writer buys shares. This is a local hedge at the current price, not a statement that all possible price movements now have zero consequences.

A trader short 100 calls on SAP, each covering one share and each with delta 0.6, has option delta −60. Holding 60 SAP shares makes combined delta zero. If each actual contract covers 100 shares instead, the required stock position is 6,000 shares. Contract-count and underlying-unit conventions must be reconciled before a hedge is executed.

Shares have zero gamma, so they cannot remove the option position’s curvature. A short conventional call has negative gamma. In an explicitly illustrative local example, a long call’s value around $S_0=50$ is approximated by $4+0.6\Delta S+\frac{1}{2}\left(0.04\right){\left(\Delta S\right)}^2$. The short-100-call portfolio plus 60 shares then has price-driven change:

A move of either +1 or −1 produces an approximate loss of 2 currency units; a move of either +5 or −5 produces an approximate loss of 50. The quadratic expression is a local approximation rather than a globally valid option price. It isolates gamma while ignoring theta, financing and volatility changes. A long option with an offsetting short stock position has the opposite local curvature.

Open figure at full size ↗

After a rise in the underlying, a call’s delta may increase from 0.6 to 0.9. The short-100-call hedger must then raise the stock holding from 60 to 90 shares. After a decline, the hedger generally sells shares as delta falls. This buy-after-rises and sell-after-falls pattern explains the rebalancing burden of short gamma. Transaction costs, market jumps, discrete hedge intervals and illiquidity prevent a dynamic hedge from being an unconditional guarantee. Delta hedging also leaves volatility, rate, basis and model risks unless these are separately managed.

Fuel options: economic protection and cash expenditure

An airline purchasing fuel is exposed to rising fuel prices. Buying calls can cap the price paid while preserving the benefit of falling prices, at the cost of an upfront premium. Consider a simplified quarterly requirement of 100 barrels, hedged by calls on exactly 100 barrels with strike USD 25 and premium USD 1.20 per barrel. Assume the options settle when the fuel is purchased, the underlying exactly matches the fuel price and there are no financing or transaction costs.

Above the strike, the option receipt offsets every additional dollar paid for the matched fuel quantity. Below the strike the option expires, but physical fuel becomes cheaper. Calling the expired option a failed hedge ignores the combined position. The premium buys protection; it is not expected to be recovered in every state of the world.

Net cash expenditure is not automatically the same as an accounting-period expense. Fuel may first enter inventory, premiums may be paid earlier, and hedge accounting may defer or reclassify components. Nor does this simple example capture real jet-fuel basis risk when the hedge references crude oil, mismatched delivery dates or uncertain fuel volumes. An airline must evaluate the matched economic exposure and the residual mismatches separately.

Time value and reported earnings: the Lufthansa example

A purchased fuel call can gain intrinsic value as fuel prices rise while losing time value as expiry approaches or implied volatility falls. The total derivative value is the sum of these components, but accounting rules may recognise their changes in different places or periods. Under a historical IAS 39 cash-flow-hedge designation that separated intrinsic value from time value, the effective designated component could be deferred in other comprehensive income, or OCI, while excluded time-value changes entered profit or loss. OCI is part of comprehensive income and equity; it is not the same as current-period net profit.

For an illustrative accounting bridge, suppose the intrinsic-value component increases by EUR 575 million while time value decreases by EUR 289 million. The derivative’s net value rises by EUR 286 million, yet a qualifying intrinsic-value-only designation could show the 575 million effective gain in OCI and the 289 million time-value loss in profit or loss, before tax and other adjustments. These illustrative amounts describe the mechanism; they should not be confused with cash settlements, total airline profit or a verified quarterly financial statement.

Lufthansa’s 2011 annual report provides a documented historical example: it reported EUR 96 million of adverse changes in option time value in the financial result, primarily from fuel hedges, while realised fuel-hedging gains reduced fuel costs by EUR 694 million. These are full-year figures. The accompanying accounting policy distinguished changes in intrinsic value recognised in equity from changes in time value recognised in the financial result. The economic assessment therefore required both the fuel bill and the hedging components, not an isolated derivative expense.

The historical treatment should not be applied mechanically to current accounts. IFRS 9 provides cost-of-hedging treatment for qualifying excluded option time value, with recognition depending on the hedge and whether it relates to a transaction or a period. Lufthansa’s first-quarter 2018 report describes a change to the recognition of option time values on adoption of IFRS 9. Entities may also elect to retain IAS 39 hedge-accounting requirements under the applicable IFRS 9 policy choice. Designation, effectiveness, documentation and the specific hedged item therefore remain essential.

The risk-management lesson is to distinguish economic cash-flow protection, derivative fair-value changes and reported earnings. A hedge can meet its economic objective while creating accounting volatility; conversely, a smooth reported earnings series does not prove that all economic exposures are hedged. Risk reports should explain the reconciliation and the periods in which hedging gains, fuel costs and option premiums affect the accounts.

Accounting and case references: Lufthansa Annual Report 2011; Lufthansa Interim Report, first quarter 2018; IFRS 9 overview; IFRS 9, paragraph 6.5.15.

What a market-risk measure is intended to describe

Market risk arises when portfolio value changes with interest rates, credit spreads, equity prices, exchange rates, commodity prices or implied volatilities. Banks, hedge funds, investment funds, pension funds, insurers, sovereign wealth funds, central banks and corporate treasuries all face versions of this problem. Their portfolios and objectives differ, but each needs to connect movements in market factors to losses that matter for its decisions. A corporate treasury may focus on foreign-currency assets or fuel hedges; a trading desk may focus on daily valuation changes; a pension fund must also consider liabilities and a longer horizon.

Different measures answer different questions. An exposure amount, such as EUR 2.5 million invested in dollar assets, describes scale but not the likelihood or severity of a euro loss. A sensitivity, such as delta, duration or the value of a basis point, describes a local response to one specified shock. Scenario analysis examines the consequences of a particular combination of shocks. Portfolio variance summarises dispersion around the mean. Value at Risk, or VaR, describes a quantile of a loss distribution; Expected Shortfall, or ES, describes the average severity of the specified tail. These measures complement rather than replace one another.

Accounting classification does not remove economic market risk. Trading positions are commonly measured at fair value through profit or loss, so price changes can affect reported earnings immediately. Banking-book assets and liabilities may be measured at amortised cost or under other accounting categories, while their economic values remain sensitive to rates and spreads. Analysis of interest-rate risk in the banking book consequently includes both economic-value and earnings perspectives. It is inaccurate to identify the entire trading book with one accounting rule or the entire banking book with amortised cost.

VaR as a loss quantile, not a maximum possible loss

Fix a valuation date, portfolio, reporting currency and risk horizon $h$. Let ${\Pi }_h$ be profit or loss over that horizon and define positive loss as $L_h=-{\Pi }_h$. With an unchanged portfolio and explicit cash flows, one possible definition is ${\Pi }_h=V_{t+h}-V_t+C{F}_h$, where $C{F}_h$ contains relevant receipts net of payments. A static instantaneous-shock calculation may deliberately exclude future cash flows and time passage, but that is a modelling convention, not a universal property of VaR.

A one-day 99% VaR of EUR 10 million means that the model assigns at least 99% probability to a loss no greater than EUR 10 million. For a continuous distribution, the exceedance probability is 1%; with discrete outcomes it need not be exactly 1%. The measure says nothing by itself about whether an exceedance is EUR 10.1 million or EUR 100 million. Calling it the “maximum expected loss” confuses a quantile with both a worst-case bound and an expectation.

The confidence level is the chosen probability threshold for the modelled distribution. It is not a statistical confidence interval for the accuracy of the VaR estimate. A model can report 99% VaR while its parameter estimates remain highly uncertain. Nor does 99% imply only one loss event in a career: under correct independent daily modelling, about 2.5 exceedances would be expected in 250 trading days. Their occurrence is not automatically evidence that the model failed; their frequency, clustering and severity must be examined.

Ordinary VaR may use a recent historical calibration or a conditional forecast of current market conditions. Stressed VaR instead uses a stressed period or distribution appropriate to the portfolio’s exposures. Both remain quantiles. A stressed estimate need not be larger for every conceivable portfolio if the selected stress episode affects different risk factors; the choice must be justified by the actual positions. Reporting VaR without its horizon, confidence level, loss definition and calibration period makes meaningful comparison impossible.

From positions and market data to a portfolio loss distribution

A VaR system combines two streams of information. The position stream identifies contracts, quantities, contractual terms, currencies and valuation models. Mapping translates these holdings into exposures to relevant market factors. The market-data stream supplies observed prices, yield curves, spreads and volatility surfaces, from which a joint distribution of shocks is estimated or selected. Revaluing the same current portfolio under those shocks produces a distribution of hypothetical profit and loss. Quantiles and tail averages are calculated only after this aggregation.

This architecture explains why “add up each instrument’s VaR” is generally wrong. The portfolio distribution depends on which gains and losses occur together. Offsetting exposures can reduce risk, while positions with different labels can share the same underlying factor. A single bond may require several interest-rate curve points, a credit spread and a currency factor. An option additionally requires volatility and nonlinear valuation. Poor mapping can omit material risks even when the quantile calculation is numerically perfect.

Market observations must be aligned in time and expressed in consistent conventions. Mixing a percentage return with a basis-point rate change without converting sensitivities creates a unit error. Stale prices can artificially suppress measured volatility; asynchronous observations can distort estimated correlations. Position reconciliation, corporate actions, missing-data treatment, independent prices and reproducible valuation are therefore part of risk measurement, rather than administrative details outside the model.

Historical reference: MSCI, Return to RiskMetrics.

Normal quantiles and a complete VaR calculation

If horizon profit or loss is normally distributed with mean ${\mu }_{\Pi }$ and standard deviation ${\sigma }_{\Pi }$, positive loss has mean $-{\mu }_{\Pi }$ and the same standard deviation. Writing $z_{\alpha }={\Phi }^{-1}\left(\alpha \right)$ for the standard normal quantile gives:

The familiar value 1.96 bounds the upper 2.5% tail and is used in a central two-sided 95% interval. A one-sided 95% VaR uses approximately 1.645 instead. Confusing the two conventions systematically changes the reported threshold. For short horizons the expected profit is often assumed negligible, but this is an explicit approximation, not part of the definition.

For an illustrative portfolio with value EUR 1 million, horizon return standard deviation 9.7% and expected return zero, the 95% normal loss threshold is $1.645\times 9.7\%=15.9565\%$, giving approximately EUR 159,565. At 99%, using the more precise quantile 2.3263479 gives about 22.5656%, or EUR 225,656. If expected return over the same horizon were 0.1%, the 95% threshold using the rounded 1.645 quantile would fall to 15.8565%, or EUR 158,565. A change of horizon requires a new return distribution or a justified scaling rule.

These are linear-return calculations. A normal return model can assign positive probability to returns below −100%, which is unsuitable as a literal model for an unlevered limited-liability equity over some horizons. A log-return model respects positive prices but requires converting log returns into monetary losses nonlinearly. Distributional convenience should not be mistaken for a universally correct description of economic payoffs.

Worked application: a twenty-year zero-coupon bond

A zero-coupon bond paying 100 currency units in twenty years has price $B\left(y\right)=100/{(1+y)}^{20}$ when the annual effective yield is $y$. At the displayed yield $y=0.0468$, the reproducible price is approximately 40.0614. A quote of about 40.04 can reflect an underlying yield rounded before presentation; calculations should use one internally consistent set of inputs. The positive magnitude of the bond’s basis-point sensitivity is:

Assume a one-day yield change with mean zero and standard deviation 9.7 basis points. A one-basis-point yield increase produces an approximately 0.0765406-unit price loss. Under a normal yield-change model, the 95% adverse yield shock is $1.6448536\times 9.7\approx 15.9551$ basis points. The linear-normal VaR is therefore:

Full repricing at $y+0.00159551$ gives a loss of approximately 1.2019 instead. Positive bond convexity makes the linear approximation overstate this loss for a yield increase. Both calculations use the same normal yield quantile; their difference comes from valuation approximation, not from changing the confidence level.

For comparison, a separate empirical sample might produce a 95% yield-change quantile of 15.15 basis points. Applying the same PVBP gives approximately 1.1596. Rounding the PVBP to 0.08 before multiplying would instead give 1.212, so coarse intermediate rounding is inappropriate. Historical quantiles and normal thresholds need not agree because one uses observed order statistics and the other a parametric distribution. The daily 9.7-basis-point standard deviation here must also not be substituted for the weekly statistics used in a different bond illustration.

Historical simulation and empirical order statistics

Historical simulation applies past joint risk-factor changes to today’s portfolio. Identify the factors that affect current holdings, collect a common sequence of historical changes, construct one current-market scenario from each historical observation, and revalue the portfolio in every scenario. The resulting loss sequence is hypothetical performance of today’s portfolio, not simply the realised performance of a portfolio whose positions changed during the historical period.

For prices, a scenario might apply a historical percentage return to the current price; for rates, it might apply an absolute basis-point change to today’s curve. The choice must suit the factor and its possible values. Observed joint movements preserve sample dependence only if the same historical date is used across factors. Resampling each factor independently destroys that dependence and creates a different model.

Let $L_{\left(1\right)}\le \cdots \le L_{\left(N\right)}$ be the scenario losses sorted from smallest to largest. Under the empirical inverse-CDF convention:

With 500 equally weighted scenarios and 99% confidence, the quantile is observation 495 in ascending order, equivalently the sixth-largest loss. Five strictly larger observations represent 1% of the sample when there are no ties. Software that interpolates between ranks may return a different finite-sample number, so the convention must be documented rather than silently switched.

An explicitly illustrative sample of twenty losses, in EUR thousands, is $-4,-3,-2,-1,0,0,1,1,2,2,3,3,4,4,5,6,7,8,10,20$. Negative losses are gains. The 95% empirical VaR is the nineteenth observation, 10, and the worst 5% consists of the final observation, 20. This example makes clear why a quantile threshold and an average tail loss convey different information.

Open figure at full size ↗

Historical simulation does not require a specific parametric density, but it still assumes that the chosen history is informative about future shocks. It retains only the skewness, tails and dependence present in that sample. A short window may omit a crisis; a long window may mix obsolete regimes with current conditions. Structural breaks, new instruments and missing price histories can make the sample unreliable. Full revaluation captures nonlinear option responses; a historical simulation using only delta approximations does not automatically capture gamma merely because its scenarios are historical.

Variance–covariance and delta–gamma approaches

Let $x$ be a vector of risk-factor changes, with mean vector $m$ and covariance matrix $\Sigma$. Let $a_i$ be the monetary sensitivity of portfolio profit to factor $x_i$, so the first-order approximation is $\Pi \approx a^{\top }x$. For percentage-return factors, $a_i$ is a currency exposure per unit decimal return; for rate changes, it must match the rate unit. The variance is:

The diagonal elements of $\Sigma$ are factor variances; off-diagonal elements are covariances. If the factor vector is jointly multivariate normal, a linear combination is normal and the delta-normal VaR is $z_{\alpha }\sqrt{a^{\top }\Sigma a}-a^{\top }m$. Marginal normality of each factor alone is insufficient: their joint dependence must be specified. Correlations are a major source of diversification, and changes in them can materially alter the result even if each volatility stays fixed.

For an illustrative two-position portfolio, let each position have one-day profit standard deviation EUR 100,000 and correlation 0.5. Portfolio standard deviation is $\sqrt{100,{000}^2+100,{000}^2+2\left(0.5\right)100,{000}^2}\approx 173,205$. With zero expected profit, 99% normal VaR is approximately EUR 402,935. If correlation becomes 1, standard deviation rises to EUR 200,000 and VaR to EUR 465,270. The diversification benefit is conditional on dependence, not a permanent property of holding two positions.

A delta–gamma approximation adds the Hessian matrix $H$ of second derivatives: $\Pi \approx a^{\top }x+\frac{1}{2}{x}^{\top }Hx$. Even when $x$ is jointly normal, this quadratic profit is generally not normal. Its quantile may require simulation or a specialised quadratic-form method. Adding gamma and then mechanically multiplying a standard deviation by 2.3263 does not generally produce an exact 99% quantile.

The delta-normal method is fast, transparent and useful for hedge comparisons because it has a closed-form calculation without Monte Carlo sampling error. Its limitations are the linear valuation approximation and the chosen distribution, covariance estimates and factor mapping. Eliminating simulation error does not eliminate parameter uncertainty or model risk. For large option books, fast approximations should be compared with full revaluation under meaningful stresses.

Monte Carlo simulation and the source of simulation error

Monte Carlo simulation specifies a joint probability model, generates many scenarios, values the current portfolio in each and reads the desired quantile from the simulated losses. A run of 10,000 scenarios is a possible starting point, not an assurance of accurate extreme-tail estimates. The method may use normal, heavy-tailed, skewed, jump or other models where justified; normality is not inherent in simulation.

For a multivariate normal example, draw independent standard normal variables $z$, choose a matrix $B$ satisfying $B{B}^{\top }=\Sigma$, and set $x=m+Bz$. A Cholesky factor can be used when the covariance matrix is suitable. Independent draws for the components of $z$ do not imply independent simulated factors: the matrix transformation introduces the chosen dependence. Each scenario then feeds a consistent pricing model, producing a monetary portfolio loss.

Full pricing can represent nonlinear options, while path-dependent instruments require simulated paths rather than only terminal shocks. A scenario distribution and valuation model must also be kept conceptually separate: market-risk probabilities usually describe real-world future changes, while option values within each scenario may be calculated using risk-neutral pricing assumptions. Confusing the two probability roles can distort the loss distribution.

Simulation error remains because the scenario sample is finite. At 99% confidence, 10,000 draws leave only about 100 observations beyond the estimated threshold. Results can be checked across independent runs or with appropriate uncertainty estimates, and more draws may improve precision. A fixed random seed supports reproducibility but does not establish accuracy. Parameter misspecification, missing factors and wrong pricing persist even with millions of scenarios. The computational flexibility must therefore be matched by calibration, validation and clear documentation.

Volatility clustering, weighting, EWMA and GARCH

Financial return distributions often have heavier tails and asymmetry relative to a fitted normal distribution. Large absolute returns also tend to cluster: turbulent periods follow turbulent periods, and quiet periods follow quiet periods. A single unconditional standard deviation can consequently describe neither regime well. Empirical histograms and return histories are evidence about these features, not a proof that one alternative model will always forecast them correctly.

Open figure at full size ↗

Open figure at full size ↗

Historical weighting gives recent observations greater influence. For an exponentially weighted moving average, or EWMA, a zero-mean innovation ${\epsilon }_t$ updates the variance forecast as:

A smaller $\lambda$ responds faster to new observations and forgets earlier shocks more quickly. In an illustrative daily model with $\lambda =0.94$, current volatility 1% and latest innovation 3%, the next variance is $0.94{\left(0.01\right)}^2+0.06{\left(0.03\right)}^2=0.000148$. The next volatility is therefore about 1.2166%, not the weighted average of 1% and 3%. Variances, rather than standard deviations, are being updated.

Ignoring the finite-sample initialisation term, an observation $j$ periods back receives weight $(1-\lambda ){\lambda }^{j-1}$. At 0.94, the most recent twenty observations carry about 70.99% of the infinite-history weight, and the most recent fifty about 95.47%. A finite window requires consistent normalisation or retention of the initial-variance contribution. Weighted historical quantiles use cumulative scenario weights rather than the unweighted rank formula.

Open figure at full size ↗

A GARCH(1,1) model adds a constant and separates the shock and persistence coefficients:

For a finite constant unconditional variance in the usual model, $a+b<1$, giving ${\bar{\sigma }}^2=\omega /(1-a-b)$. With illustrative values $\omega =0.000002$, $a=0.08$, $b=0.90$, current variance 0.0001 and innovation 0.03, the next variance is 0.000164 and volatility about 1.2806%. The long-run variance is 0.0001. EWMA corresponds to the special zero-intercept recursion with shock and persistence coefficients summing to one, so it does not have this finite mean-reverting variance formula.

These methods forecast conditional volatility; they do not automatically remove skewness, jumps or fat tails from standardised innovations. Filtered historical simulation rescales historical standardised shocks by current volatility forecasts, with dependence and multivariate rescaling treated consistently. A model that adapts rapidly to volatility can still underestimate a new regime, and its forecasts should be compared with stresses that are not mechanically limited to recent experience.

Tail severity, Expected Shortfall and diversification

VaR can remain nearly unchanged when a loss beyond the chosen quantile becomes much larger. Selling a far out-of-the-money put illustrates the problem: most days may show small premium income or modest valuation changes, while a collapse of the underlying creates a severe loss. The premium is compensation for an exposure, not evidence that the exposure has disappeared. A sample without a crash can report very low VaR for a strategy whose economic tail is substantial.

For an integrable loss distribution, Expected Shortfall averages quantiles in the worst $1-\alpha$ fraction:

With a continuous distribution this equals $E[L\mid L>{VaR}_{\alpha }]$ under the usual tail conditions. For a discrete distribution with probability mass at the VaR threshold, the simple conditional expectation can be wrong: the quantile-average definition includes the appropriate fraction of the threshold mass. In the twenty-scenario illustration, 95% VaR is 10 thousand and ES is 20 thousand because the worst 5% is one observation.

For normal loss $L\sim N({\mu }_L,{\sigma }_L^2)$, ${ES}_{\alpha }={\mu }_L+{\sigma }_L\varphi \left(z_{\alpha }\right)/(1-\alpha )$, where $\varphi$ is the standard normal density. With mean zero and unit standard deviation, 99% VaR is about 2.3263 and ES about 2.6652. ES describes severity beyond the quantile; it is still model-dependent and is not a worst-case bound.

Open figure at full size ↗

VaR is not subadditive for every loss distribution. Consider two independent positions, each losing 100 with probability 4% and zero otherwise. Each has 95% VaR zero because a zero loss occurs with probability 96%. Together, the probability of no loss is ${0.96}^2=92.16\%$, so the combined 95% VaR is 100, exceeding the sum of the standalone VaRs. This counterexample does not imply that diversification always increases VaR; it shows that VaR does not universally reward it. ES is subadditive for integrable losses and, together with monotonicity, positive homogeneity and cash-translation behaviour, is a coherent risk measure.

Quiet samples, carry trades and the Niederhoffer case

A carry trade borrows in a low-interest-rate currency and invests in a higher-interest-rate currency without fully eliminating exchange-rate exposure. An illustrative USD-funded Thai-baht investment earning 12% while funding costs 6% appears to earn a 6-percentage-point spread if the exchange rate is unchanged. The interest differential is not a risk-free arbitrage. If the baht value in dollars falls 20% over the year, one borrowed dollar produces $1.12\times 0.8=0.896$ dollars of investment proceeds but requires 1.06 dollars of repayment, a loss of 0.164 dollars before other costs.

Thailand allowed the baht to float on 2 July 1997, after pressure on its exchange-rate arrangement. This date must be distinguished from the sharp US equity-market fall on 27 October 1997 and from the timing of individual fund losses. A previously stable exchange rate can produce a low historical volatility estimate until the regime changes. A narrow calibration window may therefore mistake the absence of a past break for the absence of future currency risk.

Open figure at full size ↗

Victor Niederhoffer’s 1997 fund collapse provides a related warning about strategies with strong historical returns and severe exposure to exceptional market moves. It should not be reduced to a single elementary currency carry calculation: option exposures, leverage, market movements and funding pressures can interact. A high historical Sharpe ratio, which compares excess return with volatility, may be uninformative about rare losses when the observation window contains mainly premium income or a stable regime.

Open figure at full size ↗

The general problem is model risk at several levels: volatility and correlation estimates are uncertain; the assumed distribution may miss jumps or tail dependence; and valuation may be inaccurate precisely in stressed markets. Sensitivity to alternative windows, distributions and correlations should be visible. Stress testing asks how a plausible severe scenario would affect the portfolio even if the fitted probability model assigns it little weight. Reverse stress testing starts from an unacceptable loss and asks which combinations of events could produce it.

Historical source: IMF, Thailand stand-by credit release (1997).

Backtesting, implementation controls and the London Whale

Backtesting compares a forecast made using information available at the time with the subsequent outcome. For a 99% daily VaR, the exception indicator is $I_t=\mathrm{𝟏}\{L_t>{VaR}_{0.99,t}\}$. A well-calibrated model should have an appropriate exception rate, but frequency alone is insufficient: exceptions clustered during crises reveal conditional misspecification even if the long-run count looks acceptable. Actual trading profit and hypothetical profit on an unchanged portfolio answer different validation questions because new trades, fees and intraday decisions can affect realised results.

The London Whale episode at JPMorgan in 2012 illustrates why implementation and governance matter alongside the statistical formula. The US Senate investigation reported that a new CIO model reduced measured VaR by approximately 50% on introduction, while the positions had not thereby become safer. It identified manual operations, calculation and implementation errors and weaknesses in model review. The broader case also involved risk-limit responses and valuation controls. Attributing the entire loss to a single typographical error would obscure these interacting failures.

Open figure at full size ↗

Independent validation should examine whether mathematical specifications are implemented faithfully, compare old and new models on the same positions, reconcile input data and review abrupt risk-number reductions. A lower estimate is not itself evidence of an improved model. Material model changes need a documented economic explanation, independent challenge and clear treatment of existing limits. Sensitivity analysis, stress results and position concentrations should remain available even when a preferred VaR model produces a reassuring number.

Investigation: US Senate, JPMorgan Chase Whale Trades report (2013).

Time horizons, liquidity and the evolution of market-risk regulation

If daily profit increments are independent and identically distributed with mean ${\mu }_1$ and variance ${\sigma }_1^2$, their sum over $h$ days has mean $h{\mu }_1$ and variance $h{\sigma }_1^2$. With normal increments and stable exposures:

When the mean is negligible, ten-day normal VaR is approximately $\sqrt{10}$ times one-day VaR, not ten times. Independence and identical distributions alone establish variance scaling when variance exists; they do not make every distribution’s quantile scale exactly with the square root of time. Serial dependence, changing volatility, nonlinear valuation, jumps, rebalancing and liquidation costs can invalidate the shortcut. A position that cannot be sold quickly also remains exposed beyond a nominal one-day reporting horizon.

J.P. Morgan publicly launched RiskMetrics in 1994. The Basel Committee’s 1996 Market Risk Amendment introduced a framework permitting standardised market-risk capital calculations and approved internal models. Post-crisis reforms included the 2009 Basel 2.5 changes and stressed VaR. The Fundamental Review of the Trading Book, or FRTB, subsequently centred the internal-model framework on stressed ES and liquidity risk. The ECB’s Targeted Review of Internal Models, or TRIM, is a separate supervisory review initiative; interest-rate risk in the banking book also has its own supervisory framework rather than simply being absorbed into trading-book VaR.

The Basel market-risk standard uses 97.5% one-tailed ES and a ten-day base horizon, with specified liquidity-horizon categories extending to 120 days. The following mapping summarises the risk-factor categories in that framework; it describes the standard, not a claim that every jurisdiction applies every provision on an identical implementation date. IG denotes investment grade and HY high yield.

The specified interest-rate currencies include EUR, USD, GBP, AUD, JPY, SEK and CAD, together with the bank’s domestic currency. Exact FX-pair eligibility and supervisory adjustments require the framework’s definitions. Liquidity-adjusted ES aggregates exposures assigned to different horizons; it is not obtained by multiplying the entire portfolio by its longest horizon. Modelling the availability and quality of market observations is also necessary, because stale or sparse prices can make apparently precise factors unsuitable for the normal modelling process.

Regulatory references: Basel Framework MAR33; Basel Committee, market-risk capital requirements (2019).

Model heterogeneity and the limits of a single number

Even when institutions value the same positions, different data windows, dependence assumptions, pricing approximations and implementation choices can generate materially different capital estimates. The Basel Committee’s 2013 regulatory consistency assessment reported the following results for a common diversified test portfolio. The figures are historical model-comparison evidence, not current capital requirements for an investor’s portfolio.

The maximum is more than twice the minimum. This dispersion is a reason to investigate model choices, not proof that the smallest or largest estimate is automatically correct. Regulatory capital incentives can create pressure to favour low measured risk, making independent governance especially important. Harmonised definitions, validation, benchmarking and credible standardised alternatives help constrain unjustified differences.

A useful risk report therefore places VaR and ES beside stress losses, sensitivities, liquidity assumptions, concentrations and model limitations. VaR supplies a common threshold language; ES adds tail severity; neither removes the need to understand the positions. The quality of market data, valuation, implementation and escalation determines whether the resulting numbers support decisions or merely give false precision.

Historical comparison: Basel Committee, analysis of market-risk weighted assets (2013).

Default, deterioration and the definition of loss

A default can involve insolvency, a material payment that remains overdue, or a distressed restructuring that reduces the creditor's economic claim. Regulatory definitions also consider whether the borrower is unlikely to pay without enforcement of collateral. A 90-day past-due trigger is widely used, but its precise application depends on the applicable framework, materiality thresholds and exposure category. Legal bankruptcy is consequently neither the only default event nor a necessary condition for every credit loss. A rating of D or a non-performing-loan designation communicates a default-related status, whereas a performing loan can already have substantial credit risk.

Credit deterioration matters before default. A rating downgrade, an increase in an internally estimated probability of default or a widening market credit spread can reduce the present value of a claim while all payments remain current. Default risk, migration risk and spread risk overlap economically but are not interchangeable measurement categories. A trading portfolio may recognise spread changes through market-risk measures, while a banking-book portfolio uses default and impairment measures. Calling an exposure “credit risk” does not by itself determine its accounting treatment or its prudential classification.

Measurement must therefore specify the horizon, valuation basis and aggregation level. The eventual unpaid amount on one loan is different from the present market-value loss on a bond, and neither is the same as the unexpected loss of a diversified portfolio. Internal ratings and external ratings describe credit quality; expected loss describes a mean monetary shortfall; economic capital describes a tail buffer; and risk-weighted assets are a regulatory denominator used to determine capital requirements. Treating these quantities as substitutes obscures the decision each one supports.

PD, LGD and EAD

The three central components are probability of default, or PD; exposure at default, or EAD; and loss given default, or LGD. PD is the probability that a specified default event occurs over a specified horizon. EAD is the amount exposed when default occurs, including relevant drawings or accrued amounts rather than necessarily today's outstanding principal. LGD is the fraction of that exposure lost after allowing for recoveries and the associated costs and timing. If recovery is immediate, costless and measured on the same exposure basis, LGD equals one minus the recovery rate.

For deterministic EAD and an appropriately defined conditional mean LGD, the familiar one-period relation is

PD and LGD enter as decimals, and EAD is a currency amount, so expected loss has the same currency unit. Consider a €100 loan with a 10% probability of default and an expected recovery of €50 if default occurs. The conditional loss is €50, LGD is 50%, and expected loss is $0.10\times 100\times 0.50=5$ euros. The borrower does not pay an “expected” €5 loss in every outcome: the simplified distribution has a €50 loss with probability 10% and no loss with probability 90%.

The factorisation needs care when exposure and recovery are random. With a default indicator $I$, the general object is $E\left[IEADLGD\right]$, not automatically $E\left[I\right]E\left[EAD\right]E\left[LGD\right]$. A borrower may draw unused credit immediately before default, while collateral values fall during the same downturn. A revolving facility with €60 drawn and €40 undrawn, for example, has EAD of €80 if the assumed conversion factor on the undrawn portion is 50%. That is a modelling assumption, not a universal conversion factor. Legal enforceability, collateral liquidity and recovery costs can make LGD materially higher than the apparent shortfall between loan principal and today's collateral valuation.

For a multi-year exposure, losses at different dates and in different economic scenarios must be combined without counting the same default repeatedly. Let $q_{s,t}$ be the marginal probability of first default in year $t$ under scenario $s$, $w_s$ the scenario probability and $D{F}_{s,t}$ a discount factor consistent with the measurement purpose. A simplified representation is

The scenario weights sum to one. The loss severity should incorporate the timing of recoveries consistently, so discounting is not applied twice. Accounting expected credit loss, a market price and prudential expected loss can use different probability concepts, horizons and prescribed inputs. Sharing the labels PD, LGD and EAD does not make the resulting estimates identical.

Impairment, write-offs and fair-value losses

Under the general IFRS 9 impairment model, Stage 1 ordinarily recognises 12-month expected credit losses. This is the lifetime cash shortfall associated with defaults that could occur in the next 12 months, not merely cash payments missed during those months. A significant increase in credit risk moves an exposure to Stage 2 and lifetime expected credit losses. Stage 3 concerns credit-impaired assets and also uses lifetime expected losses; interest revenue is generally calculated on the net carrying amount. Special approaches and exceptions exist, so these three stages describe the general model rather than every financial asset.

An impairment charge changes the allowance and profit or loss; it does not necessarily mean that the borrower has already defaulted. A write-off directly reduces the gross carrying amount when there is no reasonable expectation of recovering all or part of the asset. It need not wait for the formal end of insolvency proceedings. If the loss was already fully provided for, the later write-off mainly uses the existing allowance rather than creating the same expense a second time. Changes in fair value are a separate channel for instruments measured at fair value.

The US current expected credit losses model, CECL, belongs to US GAAP and is not another name for IFRS 9. For financial assets within its amortised-cost scope, it generally recognises lifetime expected credit losses from initial recognition. Both systems are forward-looking, but their recognition frameworks differ. Comparing banks across accounting systems therefore requires more than comparing the words “expected loss” in their reports.

Accounting references: ESRB comparison of European and US expected-loss approaches; IFRS 9 impairment review, including write-offs; Federal Reserve CECL guidance.

Expected loss, unexpected loss and economic capital

At portfolio level, loss is a random variable $L$. Its mean $EL=E\left[L\right]$ describes expected loss; a quantile describes a tail outcome. Define $Va{R}_{\alpha }$ as the smallest loss level reached with cumulative probability at least $\alpha$. One common economic-capital convention is $E{C}_{\alpha }=Va{R}_{\alpha }-EL$, the tail loss above the mean. In this manual, unexpected loss is the standard deviation $UL=\sigma \left(L\right)$. Some institutions use that term for quantile-minus-mean instead; stating the convention is essential because these quantities differ.

Expected credit losses are ordinarily considered in pricing, provisions and operating income. Capital provides capacity to absorb adverse deviations and other risks. The economic-capital confidence level is a management choice linked to the institution's solvency objective; 99%, 99.9% and 99.98% are not interchangeable universal standards. Regulatory capital is determined by a separate set of rules. Risk-weighted assets, or RWA, are neither cash losses nor capital itself: a capital ratio applies a required percentage to the RWA denominator.

Consider an illustrative portfolio of 100 independent loans, each with EAD €1 million, one-year PD 2% and deterministic LGD 45%. The number of defaults $N_D$ is binomial with parameters 100 and 0.02, and total loss is $L=0.45N_D$ million euros. Expected loss is €0.9 million. The standard deviation is $UL=0.45\sqrt{100\times 0.02\times 0.98}=0.63$ million euros. The exact 99.9% loss quantile is seven defaults, or €3.15 million; economic capital is therefore €3.15 million minus €0.9 million, giving €2.25 million. The quantile is determined from the binomial cumulative distribution, not a normal approximation. This stylised example isolates diversification across independent borrowers; real portfolios can have much more severe tails because defaults are dependent.

Open figure at full size ↗

Credit risk frequently dominates banks' risk allocations, but the share depends on business model and measurement basis. Historical illustrations report credit risk as 49.6% of €77 billion of economic capital for five large German banks, with approximately €38 billion allocated to credit risk; total credit exposure was €1,938 billion and retail and small businesses represented 29.3%. Another historical savings-bank illustration allocates approximately 78% to credit risk, 20% to market risk and 1% to operational risk, with rounding. Such figures illustrate materiality rather than providing a current banking-sector benchmark, and economic-capital allocations must not be relabelled as RWA.

The economic logic of the IRB capital formula

The internal ratings-based approach, IRB, links regulatory credit capital to PD, LGD, EAD, maturity and a prescribed asset-correlation function. Its underlying asymptotic single-factor model assumes a highly granular portfolio: individual borrower shocks diversify, but a common systematic factor remains. A severe state of that factor raises the conditional default rate. For an ordinary non-defaulted corporate exposure, the core expression can be written

Here $\Phi$ is the standard normal cumulative distribution function, ${\Phi }^{-1}$ its inverse, $R$ the prescribed asset correlation, $M$ effective maturity in years, and $K$ a capital ratio before multiplying by EAD. For the basic corporate correlation function, before applicable adjustments,

The square root on the systematic-factor loading is essential. The term $LGD{q}_{0.999}$ is a conditional stressed loss rate; subtracting $PDLGD$ removes expected loss from the core unexpected-loss component. The factor 12.5 converts capital into RWA at an 8% reference capital ratio. RWA are consequently not equal to the 99.9% portfolio loss quantile. The maturity adjustment changes the charge according to effective maturity; the ordinary framework commonly constrains maturity between one and five years, with specific exceptions. Asset classes, smaller-company adjustments and certain financial-institution exposures require their own prescribed treatment.

The legacy Basel II scaling multiplier 1.06 has been removed from the revised Basel framework. Likewise, the old 0.03% PD floor is not a universal current rule: the revised framework generally uses 0.05% for corporate and bank exposures, with sovereign exceptions and distinct retail provisions. A zero observed default count does not establish zero credit risk. Long-run calibration, data requirements, conservatism and applicable floors address sparse-default samples.

Basel standards are implemented through jurisdiction-specific law. In the EU, CRR3 introduced the final Basel III package from 2025, including an output floor phased from 50% towards 72.5% in 2030. This limits the reduction in aggregate capital requirements obtainable from internal models relative to standardised calculations; it is not a floor on each individual loan's PD. Sovereign exposures are not universally assigned zero risk weight: preferential treatment depends on the applicable approach and conditions, including relevant domestic-currency provisions. Regulatory zero weight would in any event not mean economic impossibility of default.

Prudential references: Basel CRE31 risk-weight functions; CRE32 risk components; removal of the 1.06 multiplier; Regulation (EU) 2024/1623; EU output-floor transition.

Market-implied default probabilities

Credit assessment can use financial statements, management information, payment histories, external ratings and market prices. Bond and CDS prices provide forward-looking information, but they price risk under a risk-neutral measure. A risk-neutral PD is a valuation input that reproduces market prices under appropriate discounting; it is not automatically the physical frequency with which comparable borrowers will default.

Consider one-year zero-coupon bonds with face value €100. The risk-free yield is 5%, giving a price of $100/1.05=95.2381$ euros. A risky bond has a promised yield of 10%, giving a price of $100/1.10=90.9091$ euros. Suppose its maturity payoff is €100 without default and €50 after default. With risk-neutral default probability $q$,

The promised 10% yield is not the expected realised return. Risk-neutral expected payoffs discounted at 5% reproduce the price. The physical PD may be lower when investors require compensation for systematic default risk, but there is no universal fixed difference between the two probabilities. Liquidity, risk aversion, taxes, contractual features and market frictions can also affect spreads, so a simple spread-to-PD conversion can misattribute non-default components to default risk.

Historical bond spreads widen markedly in downturns, especially for speculative-grade debt. That movement can reflect worse expected cash flows, lower recovery prospects, a higher price of bearing risk and deteriorating market liquidity at the same time. A widening spread is therefore a warning signal, not a direct observation of a uniquely identified physical PD.

Open figure at full size ↗

Structural credit risk: the Merton model

A structural model connects default to the value of the borrower's assets. In the simplest Merton setting, the firm has asset value $V_t$ and one zero-coupon debt obligation with face value $F$ due at time $T$. Default occurs at maturity when $V_T<F$. With limited liability, equity receives $max(V_T-F,0)$, so equity is a call option on the firm's assets. Debt receives $min(V_T,F)$, equivalently a risk-free payment of $F$ minus a put-option payoff.

Under risk-neutral geometric Brownian motion, constant asset volatility ${\sigma }_V$, a constant continuously compounded risk-free rate $r$ and no asset payouts, equity value is

The physical default probability replaces the risk-neutral drift $r$ with the physical expected asset return $\mu$ in the distance-to-default expression. Higher leverage reduces the distance between asset value and the debt threshold; higher asset volatility increases exposure to downside states. Because corporate assets are not usually traded directly, implementation must infer asset value and volatility from observed securities and balance-sheet information. The assumption of default only at a single maturity is restrictive: real firms have several debt maturities, covenants, liquidity constraints and possible early default.

Open figure at full size ↗

Structural models and statistical accounting-score models should not be confused. A market-based expected-default-frequency approach can use structural intuition and empirical calibration, while a private-firm score based on accounting ratios estimates risk statistically. The product name or the rating provider does not establish which model is being used. The relevant questions are its inputs, target event, horizon, probability measure and validation performance.

External ratings and their meaning

External ratings order issuers or obligations by creditworthiness. The investment-grade boundary is BBB− for S&P and Fitch and Baa3 for Moody's; lower categories are speculative grade. A notch is one step within this ordered scale. An issuer rating and an issue rating can differ because a particular claim has collateral, guarantees or subordination. Rating symbols also have agency-specific definitions; a cross-agency correspondence is a broad comparison, not a guarantee that the underlying risks or methodologies are identical.

At the bottom of the scale, S&P uses default-related designations including D and SD, and Fitch distinguishes D and RD. Moody's does not use D as an ordinary long-term rating category; its C category can include obligations in default. Rating methodologies also differ in the relative emphasis placed on default likelihood, loss severity and the credit characteristics of particular obligations. “Through the cycle” describes an effort to look beyond temporary conditions, not a promise that a rating never changes during a cycle or corresponds to one permanently fixed one-year PD.

From business and financial risk to an anchor rating

A corporate rating combines business risk with financial risk. Business risk reflects country and industry conditions and the firm's competitive position. Financial risk draws on cash generation, debt burden and coverage ratios after accounting adjustments that improve comparability. An initial anchor is then modified for diversification, capital structure, financial policy, liquidity, management and governance, and comparable-rating analysis. Group or government support can change the final issuer assessment relative to the stand-alone credit profile.

In the historical process example, country risk is scored 2 (low), industry risk 4 (moderately high), their combined country-and-industry assessment is 4, and competitive position is 3 (satisfactory). The resulting business-risk profile is 3 (satisfactory). These are category scores within that methodology, not probabilities or quantities to average mechanically. The financial assessment and later modifiers are still needed before determining an issuer rating.

A historical Daimler illustration shows why reported debt is not necessarily the debt figure used in credit analysis. Operating leases and post-retirement obligations increase adjusted obligations, while surplus cash and financing-business adjustments can reduce the industrial debt measure. The following historical figures are in EUR millions. A denotes actual data; P and E denote projected or estimated columns. Repeated projected values are retained as stated, not interpreted as realised later outcomes.

The adjusted debt amount is approximately €11,258 million for 2012 and €7,457 million for the later columns. Small reconciliation differences can arise from the rounding of the individual displayed adjustments. The accounting perimeter matters: separating captive finance debt from industrial debt does not make those finance-company obligations disappear from the consolidated group or remove their economic risks.

The next historical grid translates ratios into financial-risk categories. FFO is funds from operations, EBITDA is earnings before interest, tax, depreciation and amortisation, OCF is operating cash flow, FOCF is free operating cash flow, and DCF here means discretionary cash flow, not discounted cash flow. The ratios are agency-adjusted measures. Thresholds depend on methodology and industry; this 2014-era grid explains the rating process and is not a current universal rulebook.

Combining business and financial assessments gives the following historical anchor matrix. Lower-case symbols identify intermediate analytical anchors. A satisfactory business profile with minimal financial risk produces an a/a− anchor in this example; weak liquidity or an aggressive financial policy can still alter the eventual rating. A matrix cannot substitute for assessment of the definitions underlying its rows and columns.

Default probabilities across horizons

A rating becomes more useful for measurement when it is connected to a clearly defined default frequency and horizon. The historical table below reports average cumulative corporate default rates by initial rating, measured in percent, over the 1981–2014 observation period. A cumulative rate includes defaults occurring at any time before the stated horizon. The rows are historical averages for different initial-rating cohorts; they are not forecasts for a current issuer and should not be combined as though every category had the same number of observations.