Skip to content

Risk limit

An approved boundary on a defined risk metric. Read VaR utilisation and headroom alongside measurement basis, warning thresholds, authority and escalation.

A risk limit is an approved boundary on a specified measure of exposure or risk. It turns a broad willingness to take risk into a checkable constraint: a desk may have to keep a loss estimate below a ceiling, remain within a signed exposure range, or maintain a liquidity measure above a floor. For a quant, producing the number is only part of the job; the number must be compared with the right boundary, on the right basis, in time for someone to act.
Risk appetite describes the level and types of risk an organisation is willing to accept. Limits translate that intention into operational boundaries for portfolios, business lines, legal entities or particular exposures. The FSB's risk appetite principles emphasise measurable limits, regular monitoring and institution-specific responsibilities; they do not prescribe one trading threshold for every firm.

Read the number and its rule together

Consider a fictional desk whose approved one-day 95% Value-at-Risk limit is USD 1 million. At the stated daily close, its measured VaR is USD 0.8 million. Both figures refer to the same portfolio, USD reporting currency, approved model, horizon and confidence level.

For this positive upper limit,

Utilisation=measured risklimit=0.81.0=80%,Headroom=1.0−0.8=0.2 million USD.\text{Utilisation}=\frac{\text{measured risk}}{\text{limit}} =\frac{0.8}{1.0}=80\%, \qquad \text{Headroom}=1.0-0.8=0.2\text{ million USD}.
The desk has used 80% of its VaR allowance. Its USD 0.2 million headroom is the remaining allowance in this metric. It is neither spare cash nor an amount of extra position notional that can automatically be traded. The VaR effect of a proposed trade depends on its exposures and interaction with the existing book.
Fictional one-day 95 percent VaR measured at USD 0.8 million against a USD 1 million limit: 80 percent utilisation and USD 0.2 million headroom. A separate warning begins at USD 0.9 million.
Fictional one-day 95 percent VaR measured at USD 0.8 million against a USD 1 million limit: 80 percent utilisation and USD 0.2 million headroom. A separate warning begins at USD 0.9 million.
Original fictional policy illustration, in USD millions. The dashed warning line is at 90% utilisation; the solid ceiling is the approved limit. This is a comparison of a metric with policy thresholds, not a loss distribution.

Here is the rule behind that snapshot. These choices belong to the fictional firm; another organisation may use different approvals and responses.

Policy fieldFictional desk rule
Scope and measurement basisAll positions in this desk's book, valued at the daily close in USD using the approved one-day 95% VaR model. Additional intraday checks follow material position changes.
CeilingMeasured VaR must be no greater than USD 1 million; equality is permitted by this policy.
WarningAt or above USD 0.9 million, including equality, notify the desk head and independent market-risk manager for review before adding risk. This warning does not itself change the ceiling.
Approval authorityThe risk committee approves the limit and any permanent change. The desk cannot raise its own allowance.
Monitoring responsibilityIndependent market risk checks positions and model inputs, records utilisation and reports exceptions. The desk head remains responsible for operating within the mandate.
EscalationAbove the ceiling, record a limit breach, notify the head of risk promptly and suspend additions to risk pending a decision. A temporary exception needs the authority specified in the firm's delegation; otherwise it goes to the committee. Record the reason, expiry and corrective action.

The current USD 0.8 million reading is below both the warning and the ceiling. That statement is narrower than “the portfolio is safe.”

A limit is not a promise about losses

The measured VaR is a model's estimate of a loss quantile. The limit is management's boundary for that estimate. Neither makes USD 1 million the maximum possible next-day loss. Losses can exceed VaR, and model or data errors can make the estimate misleading. A quiet market can lower estimated VaR without removing concentrated positions, poor market liquidity or funding needs.

This is why a VaR allowance may sit alongside stress testing, concentration controls and liquidity limits. Passing one test does not imply passing the others. Nor can desk-level VaRs simply be added and labelled the firm's VaR: aggregation depends on the joint loss behaviour, and VaR is not generally subadditive.

Limits can also bind in different directions. A minimum liquidity requirement has a floor, while a factor exposure may have both an upper and a lower bound. The utilisation formula above is for a positive upper ceiling; it should not be copied blindly to signed exposures or minimum requirements.

For the portfolio-construction side, Portfolio Constraints explains how sector, factor and mandate restrictions change implementable holdings. Its public preview introduces the lesson; the full teaching requires membership.

A profitable outcome does not erase an unauthorised breach. Likewise, formally raising a limit changes what is permitted; it does not itself reduce the portfolio's economic risk. A useful limit therefore has a defined metric, a reliable measurement process and a decision-maker who can respond when the boundary is approached or crossed.

Sources

  • Financial Stability Board, Principles for an Effective Risk Appetite Framework, 18 November 2013, Section II and Section III.3–4, printed pp. 2 and 6–10: relationship between appetite and limits, monitoring and assigned responsibilities.
  • John C. Hull, Risk Management and Financial Institutions, 4th edition, Wiley, 2015, Section 12.1, pp. 255–257; Section 12.3, p. 259; Section 12.5, pp. 260–261; and Section 28.1, pp. 577–579: VaR's horizon, confidence level, tail blindness and aggregation; clear limits, monitoring and breaches that produce profits. The desk, thresholds and workflow above are original fictional examples.
Last updated
  • risk limits
  • risk governance
  • value at risk
  • limit utilisation
← Back to the glossaryBack to top ↑