Basis risk in parametric insurance: a working definition

Basis risk is the gap between what a parametric (index-based) insurance cover pays and the loss actually suffered. A parametric cover pays on a weather number — a wind speed, a rainfall total, an irradiance level — not on your damage. Whenever the number and the damage disagree, the difference stays with the insured. That retained difference is basis risk.

It is the single most important concept in this market, and it is routinely under-measured.

The two failure modes

Basis risk arrives in two distinct ways, and they are different problems:

The miss. The event hurts you, but the index — measured a little differently, or averaged over a slightly wider area — never crosses its threshold. You lost money; the cover paid nothing.
The under-payment. The index fires, but pays for less of the event than you actually suffered. The cover recognises the event but not its depth.

The miss is the dramatic failure that reaches the newspapers. The under-payment is the quiet one that erodes the value of the cover every single year — and it is the one almost nobody quantifies.

Why a high correlation does not rule it out

An index can be correlated with your losses at 0.97 or 0.98 — the kind of headline number that makes a committee nod — and still leave a large share of loss unpaid. Correlation measures whether index and loss move together on average. It says almost nothing about whether they agree in the specific rare, severe events the cover exists for. A flattering correlation and a trigger that under-pays the worst events coexist happily.

"Our index correlates at 0.98" and "the cover will pay when I am actually hurt" are two different claims. Only the second one matters when the event lands. I walk through this on a concrete example in what "basis risk" really means for a wind farm.

How basis risk is measured

Not with a correlation. The honest measurement compares the index against the actual loss event by event over a long history and asks, for each damaging event: did the index fire, and did it pay in proportion to the depth of the loss? The output is a number a buyer can act on — the share of loss the cover would genuinely have transferred, and the share that would quietly have stayed behind.

Measured properly, basis risk is a property of the pairing — this trigger, on this risk — not of parametric insurance in general. The same index can be a good proxy for one portfolio and a poor one for its neighbour.

A documented example

In the 2015/16 season, Malawi's sovereign drought insurance famously paid out almost nothing for a drought that was plainly real; after a recalibration of the underlying model, the same season was worth US$8 million — and nothing about the weather had changed. Which half of that payout swing was doing the work is exactly a basis-risk question. I re-examined the episode on public data in Malawi 2016, re-examined.

Why it matters now

None of this makes parametric cover a bad product. It makes it a product that needs an honest number attached to it — a measured basis risk, for this trigger, on this risk, so the buyer knows what they are actually transferring and what they are quietly keeping. That measurement is the work I write about here: in coming pieces I put real numbers on it, using public data on European wind and solar.

This page is the reference entry of a series of notes on how well weather triggers track the losses they are meant to cover.

Disclaimer — published in a personal capacity, under my own name. This is independent work: produced entirely on my own time and my own equipment, from public information, using no resources, data, or working time of any employer, and outside my assigned duties. It carries no institutional affiliation and is neither connected with nor endorsed by any employer or institution; the views are mine alone. It is not financial, actuarial, or legal advice, and nothing here is an offer of services.