How the payout model for a parametric sandstorm policy is built

Sand and dust storms are a multi-billion-dollar peril in the Gulf, and most of it is uninsured. The World Meteorological Organization's 2025 dust bulletin puts roughly two billion tonnes of dust into the atmosphere each year, the bulk of it from North African and Middle Eastern deserts; the World Bank estimates the welfare loss across the MENA region at around USD 150 billion a year. A December 2024 winter storm shut airports, schools and public events across the Arabian Peninsula. For a date-palm cooperative, a construction contractor facing labour-law work stoppages, or a solar operator watching output collapse under an Al-Bawareh wind, the loss is real and recurring, and traditional indemnity cover is slow and expensive to settle on an event whose damage is diffuse and hard to quantify after the fact.

A parametric policy answers that by paying on a measured index rather than an assessed loss. When the index breaches a contractual threshold, the policy pays a pre-agreed amount, with no loss adjustment and no claim narrative. The whole product is the payout model: the rule that turns a weather measurement into a cash settlement. This post walks through how that model is built (index, trigger, payout curve, basis risk, pricing, settlement, reinsurance and accounting) in the order the work is done, because each step constrains the one after it. The numbers below use a fictitious insurer, "Najd Agri Mutual", and round figures chosen for clarity; a live product would calibrate them off thirty-plus years of station data.

What a parametric payout model actually does

A payout model is a function. It takes a published, independently measured environmental variable and returns a settlement amount. Everything that makes a parametric product good or bad lives inside that function: how well the chosen variable tracks the loss it is standing in for, how the threshold is set, how the payout scales, and how cleanly the settlement terms can be read off the policy when an event happens. A well-calibrated function settles quickly and cheaply, with little room for dispute. A poorly correlated index, or an over-engineered formula, pays at the wrong times and invites the disputes the structure is meant to remove. The sequence below is how the function gets built, one decision at a time.

1. Choose and define the index

The first decision is what to measure, and it sets the ceiling on everything that follows. For dust the candidates are PM10 concentration (micrograms of coarse particulate per cubic metre) at a reference station, horizontal visibility, or, for sites no ground station covers, satellite-derived aerosol optical depth (AOD) from instruments such as NASA's MODIS or the Copernicus atmospheric service. The three are physically linked: a peer-reviewed equation ties PM10 to visibility during MENA dust storms closely enough to calibrate one from the other, and high AOD readings correspond to sharply reduced visibility over the Arabian Peninsula.

The choice is not cosmetic. Ground stations measure a single point, so a farm 100 kilometres from the nearest station inherits a measurement that may not reflect its own air. Satellite AOD is spatially continuous and daily, which removes the station-placement problem for rural and remote-construction exposures, at the cost of a coarser, modelled reading. Granularity is the dominant lever on product quality: the distance between where the index is measured and where the loss happens, relative to the storm's footprint, drives correlation more than any other design choice.

Before a variable qualifies at all, it has to clear a three-part insurability gate: the index must be independent (published by a party neither side controls — in the Gulf, the WMO regional dust-forecasting centre in Jeddah is a natural authority), modelable (enough history to fit a distribution), and fortuitous (genuinely uncertain, not something either party can influence). Fail any one and the contract is not insurable as insurance — a point that comes back at the accounting step.

2. Set the trigger

The index is a continuous reading; the trigger is the rule that turns it into a payable event. The simplest trigger is a single threshold — pay if PM10 exceeds X. On its own that produces false positives, because a brief dust spike can cross a high concentration without lasting long enough to cause displacement, shutdown or yield loss. The fix practitioners use is a second condition joined by AND: the index must exceed the threshold and stay there for a minimum duration inside a defined window — for example, above 2,000 µg/m³ for 24 consecutive hours, or above 1,500 µg/m³ for at least 12 hours within 48.

The duration filter is doing real work. MENA storms that cause genuine agricultural or construction loss typically run 12 to 72 hours, not two or three; a duration gate roughly halves the false-positive rate without meaningfully dulling sensitivity to the events that actually hurt. The trigger definition also fixes the operational anchors that the settlement step depends on — the named reference station (or the rule for the nearest qualifying station within a stated radius), and the measurement window (a rolling 24-hour average, say). Those are not administrative details; they are part of the trigger, and an event either meets the stated definition or it does not.

3. Choose the payout structure

With a trigger defined, the next decision is how much the breach pays — and there are three shapes, increasing in complexity and in how tightly the payout tracks the loss.

Structure	How it pays	Basis risk	Typical buyer
Binary	Full limit if the threshold is crossed, nothing otherwise	Highest	Simple retail / small SME
Stepped	A different fixed amount per severity band	Moderate	Mid-market commercial
Linear	Scales smoothly between an attachment and an exhaustion point	Lowest	Large commercial / structured reinsurance

A binary cover is the easiest to explain and audit, but it is cliff-edged: a reading one microgram over the line pays in full, and one microgram under pays nothing. A stepped structure breaks the cliff into two or three bands — for a contractor, SAR 200,000 above 1,500 µg/m³, SAR 500,000 above 2,000, SAR 1,000,000 above 3,000 — and is where most insurance-layer SME cover sits, because it is still explainable while materially reducing the overshoot-and-undershoot problem. A linear curve removes the steps entirely with a formula — payout = min(1, max(0, (PM10 − attachment) / (exhaustion − attachment))) × limit — giving the smoothest match to loss and the lowest basis risk, at the cost of needing the index distribution integrated against the curve to price it. The structure you pick is a deliberate trade between how simple the product is to sell and how closely it tracks the thing it is protecting.

4. Quantify the basis risk

Basis risk is the gap between what the index pays and what the policyholder actually lost. It is an inherent feature of parametric cover, not a flaw in a particular design, and it runs in both directions. Take a contractor on a stepped policy whose Tier 2 pays SAR 500,000 above 2,000 µg/m³, and a storm that reads 2,100:

Scenario	Actual loss	Payout	Gap
A — under	SAR 800,000	SAR 500,000	SAR 300,000 unrecovered (a protection gap)
B — match	SAR 490,000	SAR 500,000	~zero
C — over	SAR 80,000	SAR 500,000	SAR 420,000 windfall to the insured

In Scenario A the contractor carries SAR 300,000 themselves; in Scenario C they keep a payout well above their loss. The insurer's cheque is the same in both cases because it is fixed by the index, not the loss. This two-way gap is also why parametric sits at the top of the published basis-risk hierarchy (indemnity below modelled-loss below industry-loss below parametric), and it is the honest disclosure a buyer needs before signing. The design levers that compress the downside gap are the same ones from the earlier steps: a more granular, better-correlated index, a duration-filtered trigger, and narrower bands or a linear curve. Tightening the trigger shrinks Scenario A without shrinking the coverage.

5. Price it from the index history

Parametric pricing comes from the index distribution, not from claims history — there are no claims to develop. The pure premium is a burning cost: the historical frequency at which the trigger would have fired, multiplied by the payout. For a binary date-palm cover paying SAR 500,000, with three qualifying events in thirty years of station data, the trigger probability is 10% a year and the pure premium is 10% × SAR 500,000 = SAR 50,000. Gross that up for expenses and profit — a 40% loading takes it to about SAR 83,000 — and that is the annual premium. A stepped policy is priced the same way band by band: the burning cost of each tier is its own exceedance count times its own payout, summed. A linear policy integrates the payout curve against the index distribution.

Two refinements matter at the tail. Thirty years of data hold at most one or two genuine extremes, so reading the 1-in-100-year payout straight off the raw history understates it; the annual-maximum series is fitted to a Generalised Extreme Value distribution and the price is taken from the fitted tail. And the history has to be tested for stationarity (an Augmented Dickey-Fuller test) before it is trusted, because Gulf dust frequency is trending upward with desertification and land degradation. A burning cost anchored to a 1990–2020 record can underprice 2025–2035 risk; the trend is a known adjustment, not a surprise. One discipline check sits underneath all of this: rate-on-line (the premium as a percentage of the cover limit) cannot fall below the trigger's exceedance probability, so a 1-in-100-year trigger prices at a 1% floor regardless of limit size. A structural advantage worth noting is that the risk margin on parametric is often lower than on indemnity, because there is no adverse-development uncertainty: once the index is read, the payout is known.

6. Write the settlement mechanics into the policy

A priced payout function is not yet a contract. The settlement terms have to be written into the policy in language precise enough that an event resolves itself, and their absence is the single biggest cause of parametric disputes. The policy names: the data source (the reference station ID and operator, or the satellite product and its publisher), the measurement window (for example a rolling 24-hour average), the reporting lag (the vendor publishes confirmed data within, say, three days), the settlement window (the insurer pays within 30 calendar days of trigger confirmation), and a fallback rule for data outages (if the named station is offline during the event, revert to satellite AOD at a stated conversion). These appear verbatim, not by reference, because the whole proposition is that settlement is mechanical (no adjuster, no negotiation), and that only holds if every term needed to compute the payout is fixed in advance.

7. Hedge it onward — the index flows through

When the insurer cedes the risk to a reinsurer or the catastrophe-bond market, it does so on the same index. The reinsurer does not re-underwrite the original loss; it takes on the identical trigger exceedance. The index "flows through" the structure unchanged, which keeps the reinsurance clean and fast to settle for the same reason the original policy is. The consequence to be clear-eyed about: the insurer is hedged from trigger to payout, but the basis risk (the Scenario A gap from step 4) is not transferred anywhere. It stays with the insured. The cat-bond market that absorbs this risk is large and growing (around USD 61 billion outstanding at the end of 2025, with annual issuance up sharply), so the onward capacity exists; what it does not do is close the policyholder's basis-risk gap.

8. Confirm the accounting classification

The last build step is to check what the contract is in accounting terms, because the same feature that makes parametric fast can take it outside insurance accounting altogether. IFRS 17, the financial-reporting standard for insurance contracts, applies only where the contract transfers significant insurance risk, and its definition of an insurance contract requires that the insured event adversely affect the policyholder. Paragraph B13 is explicit: a contract that pays on a specified uncertain event without requiring an adverse effect on the holder as a precondition is not an insurance contract, even when the holder uses it to hedge a real exposure — it is a derivative, accounted for under IFRS 9.

For a parametric sandstorm policy the line is drawn by one clause. If the wording requires the policyholder to have an insurable interest and to have been adversely affected — with the index breach as the trigger for, not the entirety of, the entitlement — it is an insurance contract under IFRS 17, and the payout exceeding the loss in Scenario C is fine (paragraph B12 allows payment above the financial effect). If the policy pays purely on the index with no adverse-effect precondition, the cedant likely accounts for it as a derivative. That is a reporting-and-classification question, separate again from how a solvency regime treats the contract in the insurer's capital requirement — the two layers should not be conflated. The classification does not change the payout model, but it changes which standard governs the numbers, so it belongs in the design conversation, settled with the auditor, not discovered at year-end.

What the payout model is for

The payout model is the parametric product; the policy contains little else. A traditional policy can lean on an adjuster to reconcile cover and loss after the event. A parametric policy has to settle that reconciliation in advance, written into the index, the trigger, the curve and the settlement terms, because nothing is adjusted later. That is why the build order matters: the index sets the correlation ceiling, the trigger and curve decide how much of the remaining gap is closed, the pricing falls out of the index history, and the settlement and accounting steps make the whole thing enforceable and reportable. For Gulf perils that indemnity cover serves slowly and expensively (dust on agriculture and construction, and the extreme-heat and solar-curtailment risks alongside it), a well-calibrated payout model determines whether the product pays when it should rather than at the wrong times. With the regional dust-forecasting infrastructure already in Jeddah, a regulatory sandbox open since September 2024, and Vision 2030 scaling exactly the solar and construction exposures this cover protects, the data and the demand are both in place; the difficulty is in the calibration.