Skip to content

Research

Rates and treasuries

Auction concession, pre-FOMC drift, swap-spread dislocation and curve carry include identities that were not observable as pre-registered.

Rates research here concerns mechanisms with a scheduled cause: an auction that must be absorbed, a policy meeting that must be priced, or a swap that must be funded against a deliverable. A known timetable prevents the event date from being discovered after the result.

That advantage is smaller than it looks, because the timetable itself has a history and the history is often reconstructed. A calendar that is scraped today reflects revisions, cancellations and re-scheduling that were not known at the time, and a study anchored on it is anchored on a document that did not exist when the trade would have been placed. Auditing the provenance of the schedule turned out to be a larger piece of work than testing the effect, and it is published in that order.

The treasury auction concession identity is the clearest result in this cluster and it is a negative one: it is not observable as it was pre-registered. That verdict is published in full rather than replaced by a version of the identity that happens to be measurable, because a specification adjusted until it fits the available data is no longer a pre-registration.

The pre-FOMC announcement drift is at the opposite stage. Its universe, window and pass criteria were pre-registered, and its data readiness was audited separately. Curve carry has been tested and killed. The open swap-spread work remains labeled open because the free constant-maturity series was discontinued and the modern curve is licensed.

The rates and treasuries documents