Why Market Risk Limits Must Constrain Decisions
A risk measure describes an exposure; a limit turns that measure into a constraint on decisions. For an insurer, a reported foreign-currency position or interest-rate sensitivity is not, by itself, a control. A usable limit specifies what is measured, how it is measured, the permitted threshold, who is accountable, and what happens when a proposed action would exceed that threshold.
This tutorial focuses on setting limits for positions, sensitivities, concentrations, and losses. Market risk measures such as VaR, stress loss, and factor sensitivities can support limit design, but each limit must match the risk being controlled. A limit that lacks a defined scope, unit, or measurement time can give decision-makers a false sense of control. The framework here is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Specify the Limit Before Applying It
A well-designed limit is more than a number. The limit statement should identify the risk factor or exposure, the covered positions, the measurement method and units, the threshold, the measurement frequency, and the owner. If a proposed trade affects a percentage-based concentration limit, for example, the policy must state which assets form the denominator and whether the trade changes that denominator.
The measurement basis also needs to be consistent with the decision being controlled. A sensitivity limit should state the factor, units of the shock, and whether the sensitivity is signed or expressed as an adverse-loss amount. A VaR limit should state its confidence level and holding period. A stress-loss limit should identify the scenario and whether the amount represents a loss to assets, surplus, or another defined measure. Limits can be expressed as dollar amounts, percentages, or risk measures, but different measures are not interchangeable.
| Limit type | Example measurement basis | Key design question |
|---|---|---|
| Position | Net foreign-currency exposure in dollars | Which positions are included and netted? |
| Concentration | One issuer's exposure as a share of invested assets | What is the denominator, and does a trade change it? |
| Sensitivity | Adverse dollar loss per basis point | Which factor and sign convention are used? |
| Loss measure | VaR or scenario loss for a defined horizon | What confidence level, horizon, and scenario apply? |
Ownership should be explicit at several levels. A portfolio manager may own compliance for a portfolio; a risk function may own independent measurement and challenge; and an authorized committee or executive may approve the limit and any exception process. These responsibilities should not be conflated. The person proposing a transaction should know which limit applies before approval, while the independent control function should be able to verify the calculation and data.
A warning threshold can prompt review while there is still room to adjust a decision. For instance, a policy could require consultation when exposure reaches 90% of a hard limit. That percentage is a governance choice, not a universal standard. Escalation actions should also be defined in advance, such as pausing approval, obtaining risk review, reducing a proposed transaction, or referring the decision to an authorized committee. An escalation path does not make a prohibited limit breach acceptable; it identifies who can make the decision and under what authority.
Calculating the Effect of a Proposed Decision
The useful question is often not “Is the portfolio within its limit now?” but “Would it remain within the limit after this decision?” The answer requires recalculating the measure on the post-transaction basis. The following examples use hypothetical insurer portfolios and policy thresholds.
Worked Example: Issuer Concentration With a Changing Denominator
A hypothetical insurer has $18 million of exposure to one issuer and $450 million of invested assets. Its policy caps issuer exposure at 4.5% of invested assets. A proposed purchase of \(x\) million dollars of that issuer's securities is funded with new cash that is also added to invested assets. Find the maximum purchase that satisfies the limit.
Method. Apply the concentration test to the post-purchase values. Since both issuer exposure and invested assets increase by \(x\), the constraint is
The denominator is positive, so multiply both sides by \(450+x\):
Subtract 18 and \(0.045x\) from both sides:
The maximum purchase is $2.36 million, rounded to the nearest cent in millions (equivalently, about $2,356,020.94). At that amount, the post-purchase issuer exposure is about $20.3560 million and invested assets are about $452.3560 million; their ratio is 4.5%, confirming the boundary.
Common mistake. Treating invested assets as fixed gives \(18+x \leq 0.045(450)=20.25\), or \(x \leq 2.25\) million. That calculation does not match the stated funding assumption: the purchase increases the denominator too. The correct test uses \(450+x\) and gives approximately $2.3560 million.
The calculation also shows why a limit's denominator and transaction funding convention belong in the policy definition. If the purchase were funded by selling other assets and total invested assets stayed at $450 million, the correct constraint would instead be \((18+x)/450 \leq 0.045\), giving \(x \leq 2.25\) million. The two answers differ because the modeled transactions differ, not because the limit itself changed.
Worked Example: Constraining a Transaction With a Stress-Loss Limit
A hypothetical insurer portfolio has an adverse-loss sensitivity of $28,000 per basis point for an upward interest-rate move. A proposed position adds $1,500 of adverse-loss sensitivity per basis point for each $1 million invested. Under the insurer's specified 60-basis-point upward stress, approximate the loss as sensitivity multiplied by the shock. The stress-loss limit is $2.4 million. What is the maximum transaction size?
Method. Let \(x\) be the transaction size in millions of dollars. The post-transaction sensitivity is \(28{,}000+1{,}500x\) dollars per basis point. The stress constraint is
Divide by 60 basis points to find the maximum permitted sensitivity:
The maximum transaction is $8.0 million. At that size, sensitivity is \(28{,}000+1{,}500(8)=40{,}000\) dollars per basis point, and the approximated stress loss is \(60(40{,}000)=\$2{,}400{,}000\), exactly at the limit.
Common mistake. The sensitivity is stated per basis point, so a 60-basis-point shock uses a multiplier of 60. Multiplying by 0.60 would treat the shock as 0.60 basis points and understate the loss by a factor of 100. This linear estimate is appropriate only under the stated sensitivity approximation; it is not a full repricing of a nonlinear position.
Using Multiple Limits Together
A proposed decision may satisfy one limit and violate another. When limits are simultaneous constraints, calculate the maximum permitted transaction under each one and use the most restrictive result. The measures need not be added together: a VaR amount and a stress loss are different quantities, each compared with its own approved threshold.
Worked Example: Identifying the Binding Risk Limit
A hypothetical insurer considers adding \(x\) million dollars of a market position. For its decision rule, the insurer specifies that the portfolio's 10-day 99% VaR increases by $0.22 million for each $1 million of the position, with no diversification credit in this calculation. Current VaR is $3.6 million and its VaR limit is $4.5 million. Separately, a defined stress loss is currently $5.2 million and increases by $0.35 million per $1 million added; the stress-loss limit is $7.0 million. Find the maximum position allowed by both limits.
Method. First apply the stated policy rule for VaR:
The VaR limit permits at most about $4.09 million. Next apply the stress-loss constraint:
The maximum satisfying both constraints is the smaller amount, approximately $4.09 million; the VaR limit is binding. At \(x=4.0909\) million, the stated VaR is $4.5 million and the stress loss is about \(5.2+0.35(4.0909)=\$6.6318\) million, below its $7.0 million limit.
Common mistake. Selecting the larger allowable amount because it leaves more room under one measure would violate the other constraint. Also, the linear VaR increment here is an explicit hypothetical policy rule, not a general property of VaR. In practice, a change in positions may alter diversification and requires measurement using the approved model and policy.
Why the Limit Formula Must Match the Risk
The calculations above rely on a basic but important principle: evaluate the post-decision exposure using the same definition as the approved limit. For a concentration ratio \(C(x)=E(x)/A(x)\), where \(E(x)\) is the covered exposure and \(A(x)\) is the stated asset denominator, the decision is allowed only if \(C(x)\leq L\), where \(L\) is the limit. Cross-multiplication is valid when \(A(x)>0\):
This equivalence explains why a changing denominator must be included in the algebra rather than held constant by habit. For sensitivity-based stress estimates, the corresponding approximation is that an adverse-loss sensitivity \(s\), measured in dollars per unit of factor movement, produces an estimated loss \(s\Delta\) for a shock of size \(\Delta\). It follows from the first-order approximation of value change by sensitivity times factor change. It is not exact when nonlinear effects are material or the shock is outside the range where the approximation is suitable.
Accordingly, policy should say when a sensitivity approximation is allowed and when full scenario repricing is required. A limit also needs a clear rule for aggregation: for example, whether exposure is measured by legal entity, portfolio, business unit, or the consolidated insurer. Without that scope, an apparently compliant position in one reporting unit could conceal a larger aggregate exposure.
Ownership, Monitoring, and Decision Escalation
A limit constrains decisions only if it is monitored at the time decisions are made and assigned to an accountable owner. A practical governance design records the approved threshold, measurement basis, data source, calculation frequency, responsible business owner, independent reviewer, and required action at the warning and hard-limit levels. Proposed transactions should be assessed against the post-transaction limit before execution or final approval.
Escalation should be proportionate to the limit and defined before a limit is approached. A warning trigger may require a documented review or senior approval for additional exposure. A projected hard-limit exceedance may require the decision to be paused, resized, or submitted to the designated authority under the insurer's approved exception process. The decision record should identify the exposure, calculation, reason for the request, duration if an exception is authorized, and monitoring arrangements. Escalation is not a substitute for a reliable measurement or a clear decision authority.
Common design failures include:
- Ambiguous units: A sensitivity “limit of 40,000” is incomplete unless the factor and unit, such as dollars per basis point, are stated.
- Unspecified denominator: A percentage concentration threshold needs a defined asset base and treatment of transactions that change it.
- Inconsistent horizons: A VaR limit and a stress-loss limit cannot be compared as though they measure the same horizon or scenario.
- Unclear ownership: A limit with no responsible owner or review function may not influence trade approval or monitoring.
- Overreliance on one measure: A position may be modest but have high stress sensitivity, or a VaR result may not capture the scenario that matters to the insurer.
Check Your Understanding
For each hypothetical decision, apply the stated measurement basis and identify the maximum transaction allowed by the applicable limit or limits.
- A $12 million issuer exposure is measured against $300 million of invested assets. The limit is 5%, and a purchase of \(x\) million increases both the exposure and invested assets. Find the maximum purchase. (Answer: approximately $3.1579 million.)
- A portfolio has adverse-loss sensitivity of $16,000 per basis point. A proposed transaction adds $2,000 per basis point per $1 million. For a 40-basis-point stress, the loss limit is $1.2 million. Find the maximum transaction using the stated linear approximation. (Answer: $7.0 million.)
- Current 10-day 99% VaR is $2.8 million and the policy rule adds $0.15 million per $1 million transacted; the VaR limit is $3.7 million. Stress loss is $4.0 million and increases by $0.25 million per $1 million; its limit is $5.2 million. Find the maximum transaction satisfying both limits and name the binding constraint. (Answer: $4.8 million; the stress-loss limit binds.)
- A portfolio has net exposure of $9.6 million in a foreign currency and a limit of 8% of $160 million in invested assets. A proposed transaction adds $1.2 million of net currency exposure per $1 million transacted, while invested assets remain unchanged. Find the maximum transaction. (Answer: approximately $2.6667 million.)
- A hypothetical portfolio's defined stress loss is $6.4 million and increases by $0.30 million per $1 million of a proposed position. The stress-loss limit is $8.5 million. Find the maximum position and state the appropriate action if a proposed amount is larger. (Answer: $7.0 million; do not approve the larger amount without the prescribed review and authorization.)