Designing Parametric and Index Triggers

Introduction

A parametric or index-based contract pays according to a defined measurement rather than by adjusting each insured loss. That can make a recovery faster and more predictable to calculate, but only if the contract specifies what is measured, where and when it is measured, and how the result translates into payment. A threshold without those details is not a complete trigger.

Trigger design starts with the exposure the contract is intended to address. For a wind-exposed portfolio, the relevant variable might be maximum sustained wind at a specified location. For drought exposure, it might be cumulative rainfall over a defined period and a weighted set of locations. The index need not equal an insured's loss: it must be a measurable, contractually defined signal whose relationship to the intended exposure is understood.

This tutorial develops the design choices and payout calculations. It is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Define the Trigger Before Setting the Threshold

Definition: A parametric trigger pays according to a specified physical parameter, such as wind speed, earthquake magnitude or water level. An index trigger pays according to a defined index, which may combine measurements from multiple sources, locations or periods. In either case, the contract must specify the variable, units, measurement location or aggregation rule, data source, observation period, applicable thresholds and payout function.

The distinction is about the defined trigger, not whether the number is produced by a single instrument. A gauge reading can be used directly as a parameter; a weighted rainfall total across several gauges is an index. The more processing an index requires, the more important it is to state the calculation and rules for missing, revised or inconsistent data.

A useful design sequence is to define the exposure and then make each measurement choice explicit. The index should be observable and independently verifiable, and its scale and direction should match the hazard: greater wind speed may increase payout, while lower rainfall may increase payout for drought. A contract should also say whether the observation is a peak, an average, a sum, or another statistic. Those choices can produce different outcomes from the same underlying event.

Design element	Question to resolve	Why it matters
Exposure variable	What measurable quantity tracks the intended hazard?	A poorly matched variable can respond to the wrong feature of an event.
Location or aggregation	Which station, grid cell, region or weighted set of sites is used?	Location and weighting affect whether the measurement reflects exposed assets.
Data source	Which named source and data version control?	Independent verification and a clear precedence rule reduce disputes.
Observation period	What start and end times, time zone and event rule apply?	A peak outside the stated window may not count, even if it occurs in the same event.
Payout function	What thresholds, direction, limit and calculation apply?	A defined function makes payment reproducible at and between thresholds.

Thresholds and Payout Functions

A common design is a capped linear payout. Let $I$ be an index that increases as the covered hazard worsens, $T$ the attachment threshold, $E$ the exhaustion threshold, and $L$ the maximum payment. Assume $E>T$, $L>0$, and a contract that pays zero at or below $T$, increases linearly between the thresholds, and is capped at $L$ at or above $E$. Its payment is

P(I)= \begin{cases} 0, & I\le T,\\ L\frac{I-T}{E-T}, & T<I<E,\\ L, & I\ge E. \end{cases}

The formula assumes the index is defined in consistent units and the payment is monotone in the index over the covered range. It does not apply unchanged when the trigger is binary, when higher readings reduce risk, or when the contract uses a different payout shape. Those cases require an explicitly defined alternative function.

The linear expression follows by interpolation. Between $T$ and $E$, write the payment as a straight line with slope $m$. The contract terms require $P(T)=0$ and $P(E)=L$, so $m=(L-0)/(E-T)$. Therefore $P(I)=m(I-T)=L(I-T)/(E-T)$ between the thresholds. The zero-payment and capped regions enforce the endpoint conditions outside that interval.

For a hazard that worsens when an index falls, such as rainfall deficit, the direction must be reversed. If payout starts when $I$ falls below $T$ and reaches $L$ at or below $E$, where $E<T$, the capped linear function is

P(I)= \begin{cases} L, & I\le E,\\ L\frac{T-I}{T-E}, & E<I<T,\\ 0, & I\ge T. \end{cases}

The thresholds should be selected with the exposure and the intended risk transfer in mind, not chosen solely because they make a convenient formula. The contract should also say whether reported values are rounded before applying thresholds or whether the calculation uses the source's full precision. If rounding is intended, the rule and number of decimal places belong in the specification.

Worked Example: A Wind-Speed Trigger With Measurement Uncertainty

Method: Apply the increasing capped linear payout to the contractually specified wind measurement, then recompute using the alternate reading to quantify sensitivity. In this hypothetical contract, the measured variable is the maximum 10-minute sustained wind speed at a named grid cell during a specified 24-hour event window. The controlling data source reports kilometres per hour. The attachment threshold is $120$ km/h, the exhaustion threshold is $180$ km/h, and the maximum payment is $\$2{,}400{,}000$.

The slope between the thresholds is

\frac{\$2{,}400{,}000}{180-120} =\$40{,}000\text{ per km/h}.

The source's finalized reading is $153$ km/h. It is above attachment and below exhaustion, so the recovery is

P(153)=\$2{,}400{,}000\frac{153-120}{180-120} =\$2{,}400{,}000\frac{33}{60} =\$1{,}320{,}000.

Suppose a permitted independent measurement, used only for sensitivity testing, is $149$ km/h. Its payment would be

P(149)=\$2{,}400{,}000\frac{149-120}{60} =\$2{,}400{,}000\frac{29}{60} =\$1{,}160{,}000.

The four km/h difference changes the indicated payment by $\$1{,}320{,}000-\$1{,}160{,}000=\$160{,}000$. The finalized contract source, not whichever reading produces the preferred result, controls the actual payment. Check: $153$ km/h lies between $120$ and $180$, and $\$1{,}320{,}000$ is between zero and the $\$2{,}400{,}000$ limit. Common mistake avoided: applying the payment limit as the slope, rather than dividing it by the 60 km/h interval.

Data Sources, Aggregation and Missing Observations

For an index built from several measurements, the contract must define how those measurements become one value. A weighted index might be $I=\sum_{j=1}^{n}w_jx_j$, where $x_j$ is a specified observation and the weights $w_j$ are fixed in advance, nonnegative and sum to one. Weights should reflect the exposure purpose of the index; they should not be adjusted after an event to change the result.

A robust data specification identifies the source and version that control, measurement units, location identifiers, observation frequency, time zone, rounding, and treatment of revisions. It should also state a fallback rule for missing or invalid readings. A fallback could name a secondary source or a defined substitution procedure, but should not leave either party free to select a replacement after seeing the outcome. If no valid value can be established under the rule, the contract must state the consequence.

Measurement error matters because it can move an index across an attachment or exhaustion threshold, or change a payment within the linear range. A useful design review tests alternative plausible readings near each threshold, not just at the expected index level. This is a sensitivity check, not permission to replace the contractually controlling data with an alternative.

Worked Example: A Weighted Rainfall Index for Drought

Method: Calculate the specified weighted index first, then apply the decreasing capped linear payout. Consider a hypothetical agricultural cover with three fixed gauges. The contract index is the weighted total of rainfall over one stated 30-day period: $I=0.50x_1+0.30x_2+0.20x_3$, with rainfall in millimetres. The finalized readings are $x_1=38$, $x_2=44$ and $x_3=52$ mm. Payment begins when the index falls below $T=60$ mm and reaches its $\$900{,}000$ limit at $E=30$ mm or less.

First calculate the weighted index:

I=0.50(38)+0.30(44)+0.20(52) =19+13.2+10.4 =42.6\text{ mm}.

Because $30<42.6<60$, the index is in the linear payout range. The payment is

P(42.6)=\$900{,}000\frac{60-42.6}{60-30} =\$900{,}000\frac{17.4}{30} =\$900{,}000(0.58) =\$522{,}000.

As a sensitivity check, suppose the first gauge were reported as $42$ mm rather than $38$ mm, with the other readings unchanged. The index would increase by $0.50(42-38)=2.0$ mm to $44.6$ mm, and the payment would be

P(44.6)=\$900{,}000\frac{60-44.6}{30} =\$900{,}000\frac{15.4}{30} =\$462{,}000.

The four-millimetre change at one gauge reduces the index payment by $\$60{,}000$, because that gauge carries half the index weight. Check: $42.6$ mm is between the 30 mm exhaustion and 60 mm attachment levels, and $\$522{,}000$ is below the $\$900{,}000$ limit. Common mistake avoided: averaging the three gauge readings equally. That would disregard the contractual weights and calculate a different index.

Trigger Timing and Event Windows

A trigger needs a time rule as well as a variable rule. The specification should define the event window's start and end, whether endpoints are inclusive, the time zone, and how observations are assigned to the window. For a peak measurement, it should specify the sampling interval and whether the trigger uses the maximum valid observation. For a cumulative measurement, it should state which observations are included in the sum.

Timing affects recovery even if the instrument is accurate. A genuine peak may fall just outside a fixed window; shifting that window can therefore change whether the trigger attaches. It is also important to distinguish the time of the physical observation from the time at which the source publishes or finalizes it. A late report does not necessarily move the underlying observation into a later event window. Settlement deadlines and rules for revised data should be stated separately from the observation rule.

Worked Example: A Water-Level Trigger and a Fixed Event Window

Method: Identify the maximum reading inside the contract window, then apply the increasing capped linear function. A hypothetical flood cover uses the maximum 15-minute water-level reading at one specified gauge from 00:00 UTC on 12 September through 00:00 UTC on 15 September, with both endpoints included. The attachment is $4.8$ metres, exhaustion is $6.3$ metres, and the payment limit is $\$1{,}500{,}000$.

The finalized readings show a maximum of $4.7$ m inside the window and a peak of $5.4$ m at 03:00 UTC on 15 September. Since the end time is inclusive, the 03:00 observation is outside the window. The contractual index is therefore $4.7$ m. Because $4.7\le 4.8$, the payment is zero.

For comparison, suppose the contract had instead defined a shifted window that included the $5.4$ m observation. Its payment would be

P(5.4)=\$1{,}500{,}000\frac{5.4-4.8}{6.3-4.8} =\$1{,}500{,}000\frac{0.6}{1.5} =\$600{,}000.

Thus the window definition changes the hypothetical recovery from zero to $\$600{,}000$, even though the physical event and gauge readings are unchanged. The shifted-window result is a comparison, not the payment under the stated fixed-window contract. Check: $5.4$ m is between the two thresholds, and $0.6/1.5=0.4$, so the payment is 40% of the limit. Common mistake avoided: using the event's peak without checking whether its timestamp falls inside the contractual observation period.

Design Review and Common Traps

Before finalizing a trigger, test the full calculation at values below attachment, exactly at attachment, between thresholds, exactly at exhaustion and above exhaustion. Confirm that the direction of payment matches the hazard, the units are consistent, and the stated limit is respected. Also test readings close to a threshold and scenarios involving missing data, revised source values and observations near the event-window boundary.

Ambiguous exposure measure: “maximum wind” is incomplete unless the contract states sustained or gust speed, the averaging interval, the location and the applicable time window.
Reversed payout direction: a drought index based on rainfall generally pays more as rainfall falls, not as it rises.
Incorrect interval: the linear formula uses $E-T$, the distance between thresholds, not the threshold $E$ alone.
Unspecified data authority: a secondary source should be used only under a defined precedence or substitution rule.
Confusing publication with observation: a report's release date does not automatically change the time assigned to the underlying measurement.

Key takeaway: A well-designed trigger is a complete rule: it defines the exposure variable, location or aggregation, units, controlling data source, observation window, thresholds, payout function and data exception procedures. The formula is only as reproducible as those definitions.

Check Your Understanding

For each hypothetical contract, apply the stated measurement and payout rules exactly.

A wind trigger pays linearly from $100$ to $160$ km/h, with a limit of $\$600{,}000$. The controlling reading is $130$ km/h. What is the payment? (Answer: $\$300{,}000$)
A drought index is $I=0.60x_1+0.40x_2$. The readings are $35$ mm and $50$ mm. Payment starts below $55$ mm and reaches a $\$500{,}000$ limit at $25$ mm. What is the index and payment? (Answer: index $41$ mm; payment $\$233{,}333.33$)
An increasing trigger has attachment $8$, exhaustion $14$, and a $\$1{,}200{,}000$ limit. The index is $15$. What is the payment, and why? (Answer: $\$1{,}200{,}000$; the index exceeds exhaustion, so the payment is capped.)
A two-site index is the equally weighted average of $72$ mm and $48$ mm. A decreasing trigger pays up to $\$800{,}000$, attaching at $70$ mm and exhausting at $40$ mm. Calculate the index and payment. (Answer: index $60$ mm; payment $\$266{,}666.67$)
A water-level contract uses the maximum reading from 00:00 through 12:00 UTC, inclusive. Its attachment is $3.0$ m, exhaustion is $5.0$ m and limit is $\$400{,}000$. Readings are $2.8$ m at 11:45 and $4.0$ m at 12:15. What is the payment? (Answer: $\$0$; the $4.0$ m reading is outside the window, and the in-window maximum is below attachment.)

What You'll Learn