Several independent estimators do not appear by themselves. Estimating the floor across many regions takes data, skill, and years of steady work. If only a government agency can afford it, there will be one estimator, and one estimator is what the system exists to avoid. So estimating has to be paid, possible for small bodies, and worth doing well.
The rule behind every incentive
Pay for doing the work well, never for agreeing.
The temptation is to reward estimates that land near the median, on the theory that the median is probably right. It is the worst incentive available. Estimators would learn to guess where the others will land and land there, and the independence that makes the median trustworthy would disappear. An estimator who sees something the others miss would be punished for it.
So the system rewards only two things: work done openly to a published standard, and accuracy proven later against what actually happened.
The measurement fund
Estimators are paid from a measurement fund, a small fixed share of the public share, set as a rule that is reviewed on a fixed cycle and hard to cut (how the guarantees change). It is small next to what it protects: an error of one percent in the floor measure moves more money each month than the fund spends in years.
The fund is held apart from the bodies that pay the floor and issue money, so that no one who benefits from a lower figure sets the estimators’ pay.
Four incentives
1. A base fee for every accredited estimate
Every estimate that meets the standards (method, data, uncertainty, independence) earns a fixed fee for the cycle. The fee is the same whatever the estimate says. It is sized to cover the cost of a small capable team, so a university center or a civic institute can afford to take part, not only large agencies.
The base fee is what makes entry possible. It pays for showing up and doing the work properly.
2. A reward for accuracy, proven later
Some time after each cycle, once the facts are in, estimates are checked against what actually happened. Estimators with strong records over several cycles earn a further reward, up to the size of the base fee again.
Accuracy is measured against reality, never against the median. An estimator who was alone and right earns the most.
How each figure is checked later:
| Figure | Checked against | When |
|---|---|---|
| The floor measure | rents and prices for floor-standard homes found in a later field audit | the following year |
| The supply gap | whether shortages appeared where the gap said: vacancies, waiting lists, rents | one to three years later |
| The care gap | unmet care and waiting times actually recorded | the following year |
| The formation gap | posts that stayed unfilled, programs that stayed out of reach | the following year |
| Sustainable output | the revised accounts, once full data on depletion is in | two to three years later |
| Concentration | ownership records once disclosed or examined | as records become available |
The audit that supplies the facts is done by a body that does not estimate and does not accredit. Its sampling method is published, and it is drawn fresh each cycle, so no estimator can know in advance which homes or which care services will be checked.
3. Access to data
Accredited estimators get access to public records they could not otherwise use, anonymized so no person can be identified: rents recorded at registration, prices, building costs, care hours, holdings reported to the measuring bodies. The data any estimator gathers for the system is shared with every other estimator after each cycle.
This lowers the cost of entering more than any fee. The hardest part of estimating is usually getting the data, not analyzing it.
4. A public track record
Every estimator’s history is published: each estimate, its later error, every change of method and the reason given. Over time reputation becomes the strongest incentive of all, and it points the right way, toward being right rather than being safe.
Who can be an estimator, and where its money comes from
An estimator’s income has three sources, and only three:
- The base fee, from the measurement fund, for every accredited estimate.
- The accuracy reward, from the same fund, paid once later facts confirm the estimate.
- Its own funding: a public agency’s budget, a university’s research funds, an institute’s members and donors, grants. Every source counts toward the cap on any one funder’s share, and the measurement fund counts as a funder too.
Any kind of body may be an estimator if it meets the standards: a public agency, a university, a nonprofit, a body in another country, or a for-profit research firm. A for-profit estimator follows the same rules as everyone else. An investor who owns or funds one is a funder or a controlling group like any other: they count toward the funding cap, several estimators they own count as one, and anyone holding a stake in what the figures affect is barred.
One rule binds every estimator and everyone who funds or works for one: no one may profit from the direction of a figure. No betting on the floor measure, no trading on a figure before it is published, no stake in a market the figures move. Income comes from doing the work and being right, never from what the number turns out to be. That leaves little for an investor to chase, which is the point: most estimators will be public agencies, universities, and nonprofits, and a for-profit firm that joins earns a modest return for good work.
How the incentives resist gaming
| Attempt | Why it fails |
|---|---|
| Copying the expected median | earns nothing extra, and correlated estimators are counted as one, so the copier loses its separate place in the median |
| Shading the figure to please a funder | funding caps limit any funder’s share; the later audit exposes the shading in the track record |
| Many small estimators run by one group, to collect fees and votes | ownership and funding are looked through; affiliated bodies count as one |
| Claiming accuracy rewards by revising old estimates | estimates are fixed and published when submitted; only the original counts |
| Gaming the audit | the audit body is separate and draws its sample fresh each cycle |
What estimators are not paid for
Estimators are never paid by the figure they produce, never by the bodies their figures affect, and never for speed at the cost of method. They are not asked to agree with anyone, and the system is designed so that disagreeing, when the evidence supports it, is the most valuable thing an estimator can do.