Saturation and Rebasing
Saturation is when a metric pins at the top of its scale and stops registering growth; rebasing resets the scale to fix it. How rebasing is governed — especially on the downside — is what separates an honest index from a tunable one.
Rail: Macro · Updated: 2026-07-09
What It Is
Saturation — known in statistics as a "ceiling effect" — happens when a normalized metric approaches or pins against the top of its fixed range. Once it saturates, the metric loses its discriminatory power: further real-world growth no longer moves the score, so genuine momentum goes unregistered and the top of the scale becomes indistinguishable. It's a recognized vulnerability of any bounded, indexed measure.
The response is rebasing (also called reconstitution or re-benchmarking): formally resetting the baseline, revising the goalposts, or updating the underlying basket so the measure stays meaningful as the phenomenon grows. National statistics agencies update Consumer Price Index base years and baskets; equity index providers like S&P and MSCI reconstitute their indices on fixed schedules; the UNDP periodically raises HDI goalposts as global life expectancy and schooling climb. Because resetting an index's parameters creates a break in the series, best practice is to document the change and use an overlap or "chain-linking" period — computing the series under both the old and new parameters at the changeover so the history stays interpretable across the break.
Crucially, because rebasing alters the calculation, it carries a real risk of manipulation if done opaquely: a reference point can be reset to flatter a stalling series. The field's defense is pre-defined, rules-based methodology and transparency, so a rebase is a predictable, disclosed event rather than a discretionary adjustment.
Why It Matters for the Machine Economy
The MEI builds this whole lifecycle in as governed, pre-committed rules rather than ad-hoc adjustments. Each metric has a saturation trigger (a review fires when a normalized score sits above 95 or below 5 persistently, not on a transient spike), and when an upward rebase happens it comes with a mandatory disclosure package that matches the statistics-agency overlap practice above: the old bounds, the new bounds, the value scored under both at changeover (so the level effect is visible), a version increment, and a period of parallel-series display. Scores stay tagged to the methodology version they were computed under and are never silently revised. One metric — Nvidia data-center revenue — is even admitted knowing it will hit its saturation trigger within a couple of years, with that early rebase documented in advance as designed behavior marking the point where that phase of the compute build-out has matured.
The part that matters most for credibility is the downward rule, because it's the strongest anti-gaming pre-commitment in the whole methodology. A floor in the MEI is a dormancy claim — the level at which the thing is effectively absent — not a growth expectation. So approaching the floor is never, by itself, grounds to lower it. Lowering a floor to chase a collapsing metric would manufacture headroom for decline and launder exactly the contraction the index exists to register. A metric drifting toward its floor rides it, flagged "distressed," and the only legitimate outcomes are a measurement-grounded correction (a source or definition genuinely changed) with full disclosure, or eventual structural retirement — never a quiet downward reset to make the number look better. This asymmetry is published, and it's what makes "the index can decline" a real property rather than a claim.
Real-World Example
When the UNDP sees a dimension like expected years of schooling approaching its ceiling, it raises the goalpost so the metric can still distinguish top performers — an upward rebase, chain-linked so history stays comparable. The MEI does the same on the upside (with its full disclosure package), but pointedly refuses the mirror-image move on the downside: it will not lower a floor to accommodate a shrinking metric, because that would hide the very decline the index is supposed to show.
Related Terms
- Normalization — the bounds that saturation and rebasing act on
- MEI (Machine Economy Index) — the index these governance rules protect
- Backcasting — reconstructing history when a metric or definition changes
- Composite Index — the general class of long-lived measure this maintains