Basis Risk Is a Data Problem Before It Is a Structuring Problem

Basis risk is the gap between what a parametric contract pays and what the protection buyer actually lost. It is the central objection to parametric cover, and the usual response is structural: narrow the trigger box, add intensity bands, layer a second trigger.

Those help. But a meaningful share of basis risk is decided before structuring begins, by properties of the measurement itself.

Three properties that set the floor

Latency. A trigger is only as timely as the feed behind it. A wind-speed trigger referencing advisory data inherits that advisory's cadence. This does not usually change whether a bond triggers, but it changes when anyone can know, and that interval is where operational risk lives — positions get marked, calls get taken, decisions get made on a stale number.

Spatial resolution. A trigger defined against a point measurement and a loss distributed across a portfolio are different geometries. The finer the reported geometry, the less of the mismatch is unavoidable. This is why a trigger written against a gauge with a known location behaves differently from one written against a regional summary, even at identical thresholds.

Revision policy. Physical measurements get revised. Earthquake magnitudes are refined as more stations report. Storm intensities are reanalysed post-season. A contract that references "the reported value" without naming a version or timestamp has imported a revision policy it never negotiated. Anyone who has watched a magnitude move by a tenth after the fact knows this is not hypothetical.

What this implies for monitoring

If those three properties set the floor on basis risk, then monitoring infrastructure should make them visible rather than hide them behind a single number.

Concretely, that means a trigger evaluation should record which source produced the measurement, when that source published it, what the value was at that moment, and what the threshold was — not merely whether the threshold was crossed. A boolean tells you the outcome. It does not let anyone reconstruct the outcome six months later when it is being questioned.

It also means being explicit about the difference between "the threshold was not reached" and "we could not evaluate the threshold". Those are different states with different consequences, and a system that reports them identically is not neutral about basis risk — it is understating it.

The honest version of the pitch

Better monitoring does not remove basis risk. Nothing outside structuring does. What it removes is a specific and avoidable component: the part that comes from not knowing, promptly and with provenance, what the reference measurement currently says.

That is a narrower claim than the category usually makes, and it is the one we would rather be held to.