There are four ways to take over a set of estimators without anyone noticing. Own several of them, so a majority answers to one party. Fund them, so they lean the way the money leans. Make them alike, so they share one method’s errors and the median inherits them. Or decide which of them count, so inconvenient estimates are quietly dropped. Each safeguard below blocks one or more of those paths, and together they leave no single party able to set the figure.
1. One controlling group, one counted estimator
Ownership and control are looked through to the people behind them, the same way the lifetime allowance looks through trusts and companies. A group that runs several estimators, under different names, counts as one. So does a funder that supplies most of several estimators’ budgets, and a network whose estimators share staff or leadership.
Blocks: owning a majority under several names.
2. Caps on any funder’s share
No funder may supply more than a set share of any estimator’s budget. The starting point is one fifth; the figure is a rule under the tier system, reviewed on a fixed cycle. The measurement fund itself counts as one funder, so even public money cannot make an estimator dependent on it alone.
No estimator may take money, in any form, from a firm or holder whose duties or limits its figures affect: a holder of scarce goods, a builder of an essential layer, an industry whose depletion it measures.
Blocks: steering by funding, and the measured paying the measurer.
3. A mix of kinds
The counted set must include different kinds of estimator:
| Kind | Example |
|---|---|
| Public agencies | a national statistics office, a regional data body |
| Universities | a research center, a joint center of several universities |
| Civic bodies | an independent institute, a cooperative research group |
| Bodies from other countries | a foreign statistics office, an international body |
No kind may form a majority of the counted set, and at least three kinds must be present. Estimators with different political leanings are welcome. Biases that compete tend to cancel; a single body that claims to have none simply hides its own.
Blocks: an estimator set that shares one institution’s blind spots, or one state’s interests.
4. Correlated estimators count as one
If two or more estimators move in lockstep, sharing the same data, the same method, and the same errors cycle after cycle, the rule treats them as a single estimate. The test is published: when the later audit shows their errors running together over several cycles beyond a set threshold, and their methods overlap, they are merged for counting until they diverge.
This catches two different problems with one rule: hidden common ownership that slipped past the look-through, and plain copying.
Blocks: making estimators alike, deliberately or by imitation.
5. A random draw each cycle
The estimates that count in each cycle are drawn at random from the accredited pool, within the mix-of-kinds requirement. The starting point is seven counted estimates drawn from a pool that should be larger. Every accredited estimate is still published, counted or not, so the public can see whether the drawn set and the full pool agree.
A random draw means no one knows in advance which estimators will set the figure. Pressuring an estimator buys a chance at influence, not a certainty, and pressuring enough of them to matter would mean capturing most of the pool, which the other safeguards make visible.
Blocks: targeted pressure on the bodies known to count.
6. A separate accreditor that judges only method
One body decides who may enter the pool. It does not estimate, does not fund estimators, and does not pay the floor. It may judge three things only: whether an estimator’s method is published and sound, whether its data is open or obtainable, and whether it is independent under the rules above. It may never judge an estimator by its results.
Every decision to admit, refuse, or remove an estimator is published with its reasons and can be appealed to the courts. Removing an estimator because its figures were inconvenient is barred outright. The accreditor’s members rotate, none comes from an estimator or a measured industry, and the accreditor itself is reviewed for drift on the same cycle as the rules it applies.
Blocks: deciding who counts by deciding who may enter.
7. A minimum number counted
The median needs at least five independent estimates, after correlated estimators are merged. With fewer, the figure is flagged as unreliable and the last sound figure stands, adjusted only for the general change in prices, until enough independent estimates return. No figure is ever set by a single estimator, even in an emergency.
Blocks: a figure set by whoever is left after the rest are driven out.
8. Cooling-off and conflict rules
The rules in when favors are for sale apply to estimators and accreditors: no gifts, holdings at arm’s length, recusal from anything touching their own interests, and a cooling-off period before staff move to a body whose figures they shaped.
Blocks: buying an estimator one person at a time.
How they work together
| Path to capture | Safeguards that block it |
|---|---|
| Owning several estimators | 1, 4 |
| Funding them | 2, 1 |
| Making them alike | 3, 4 |
| Deciding which count | 5, 6, 7 |
| Buying their people | 8, 6 |
| A government setting its own figure | 2, 3, 6, 7 |
The last row matters most, because the party with the strongest reason to want a lower floor or a smaller care gap is often the one paying them. A government may run an estimator, as any party may. It may not run two, fund more than its share of any, make up a majority of the counted set, choose which estimates count, or set a figure alone when others are missing.