Room: THE STATE. The Party’s list of names — and what a list picked for loyalty does to a ruling layer’s competence. The centre is a real, runnable adverse-selection model: merit picks the ablest, loyalty picks the most reliable, and the gap between them is measured.
source The nomenklatura — the Communist Party’s list of key posts filled by politically-approved appointees, and the privileged class that held them. Analysed by Michael Voslensky, Nomenklatura: The Soviet Ruling Class (1980). Rendered, not quoted. Ties: gobernet · the-ratchet-effect · the-soft-budget-constraint.
Nomenklatura is literally a list of names (from Latin nomenclatura). In the Soviet system it was the register of key posts — ministries, factories, universities, the army, the press — that could only be filled by candidates the Party had vetted and approved.
To hold a nomenklatura post was to enter a privileged class: special shops, dachas, foreign travel. Michael Voslensky, a defector from inside it, named it the Soviet ruling class (1980) — a self-perpetuating elite selected by the centre, for the centre.
The mechanism in one line: posts are filled by political reliability, not merit. A candidate’s value to the appointer is measured by loyalty l — will they obey, protect the patron, never threaten — more than by competence c, the ability to actually do the job.
When l and c are the same thing, no harm done. The pathology begins the moment they come apart — when the most loyal are not the most able, and the list still picks by loyalty.
This is a named idea in economics and political science, not a Soviet curiosity:
Adverse selection (Akerlof) — a selection rule optimising the wrong signal degrades the pool. Principal–agent theory — the agent (appointee) is chosen for alignment to the principal, not output. Patronage & “negative selection” — documented across autocracies and machine politics well beyond the USSR.
The mechanism outlived the state that made it famous. Any organisation that promotes for deference over ability re-runs it.
A fixed pool of 200 candidates, each with a competence c and a loyalty l, drawn once from a seeded generator (deterministic — reload and it is identical). The regime sets how loyalty relates to competence.
Two selection rules over the same pool. MERIT takes the top-k by competence. LOYALTY (the nomenklatura) takes the top-k by loyalty. Compare the mean competence of each set against the pool average.
Let each cohort of loyal appointees choose its successors by loyalty, from candidates no abler than the average incumbent (a patron rarely hires a rival). Competence ratchets down, cohort by cohort — while merit selection stays flat. The ruling layer decays in capability. LIT
The pathology is not loyalty; it is loyalty overriding competence — when the signal being optimised (obedience) is not the signal the job needs (ability), and the two have come apart. This model measures exactly that gap; it does not claim loyalty is a vice.
“Loyal cadres are the best cadres — selecting for loyalty also gets you competence.” Cut. Only true when loyalty and competence are perfectly aligned. The moment they part — and in yes-men selection they anti-align — loyalty selection forfeits competence: its mean falls below merit, and below the pool itself.
“The list is neutral — it just fills posts.” Examined. A list is a selection rule. Which signal it sorts on is the whole story; sorting on loyalty is not neutral.
Press the claim the system told itself: “loyalty selection also gets the best competence.” The witness checks it against the real pool — loyalty’s mean competence is below merit’s (and, for yes-men, below the pool) — and turns red.
The claim is falsifiable against the data. The nomenklatura trades competence for loyalty; asserting otherwise makes the witness disagree with the arithmetic.