Interest Rate Curves
Table of Contents
1. Summary
A discount curve answers one question — what is a future cashflow worth today — and everything else in this cluster exists because that question turned out to be harder than it looks. The curve is not observed. It is bootstrapped from a handful of quoted instruments, which fixes its value at a few tenors and leaves every point between them to interpolation. The 2008 crisis then broke the assumption that one curve could both discount a cashflow and project the rate that cashflow depends on, so a single curve became a family, and the family needs an order to build it in, a way to be stored, and a way to be namespaced when the same currency has more than one.
Read the cluster in that sequence: what a curve is, what it is made of, why there are several, and how the several are held. Each step creates the problem the next one answers, which is why this order and not the alphabet.
2. Reading order
2.1. What a curve is, and how one is built
Interest Rate Curves — start here. Day-count conventions, discount factors, and bootstrapping from deposits, FRAs, swaps and OIS. Everything downstream assumes this vocabulary, and the bootstrap is where the cluster's central fact appears: the curve is constructed from a small set of quotes rather than observed.
2.2. What it is made of, and what fills the gaps
Pillar — the tenor points that are genuinely quoted inputs, as against the continuum the curve pretends to be. Read it immediately after the bootstrap, because it names what the bootstrap actually produces.
Interpolation — what happens between pillars, and the local versus non-local method families. This follows pillars necessarily: interpolation is only a question once you know the curve is defined at finitely many points.
2.3. Why there is more than one curve
Multi-Curve Construction — the history and the build order: one curve served both discounting and projection until basis spreads made that untenable; now the Funding Curve is built first and each Projection Curve is discounted off it. Read the history before the taxonomy, because the taxonomy is a response to it.
Funding and Projection Curves — the resulting family: one funding curve for discounting, one projection curve per floating index tenor, and why curve identity is a discrete label rather than a continuous axis. That last point is the one most often got wrong when modelling curves as data.
2.4. How the family is held
Interest Rate Curve Bundles — how a currency's curves compose above the level of a single curve.
Curve Sets Storage — the storage-only grouping of several small related curve time series into one composite. Storage-only is the load-bearing word: it is a persistence concern and carries no pricing meaning, and conflating it with a bundle is a recurring mistake.
Onshore/Offshore Curve Namespacing — when one currency has parallel discount curves selected by booking jurisdiction. Read last: it is a refinement of identity, and only makes sense once curve identity is understood as a discrete label.
2.5. Alongside, when the question is provenance
Curve Point Provenance — where a curve point came from and why market data needs lineage at all. Not part of the construction story, so it does not sit in the sequence; reach for it when the question is trust rather than shape.
3. What this cluster does not cover
Volatility surfaces are a separate cluster with their own structure note, even though both are term structures built from quoted pillars. The resolution of a logical market data need into the physical identifiers a curve is built from belongs to Market Data Requirements and Resolution.
4. See also
- Knowledge — the index this structure note hangs from.
- Zettelkasten — what a structure note is and why the order is authored rather than computed.