Net sequestration capacity of ₽1.5–2.5 trn annualised is the standard number, and it tells you nothing about which item will actually be cut, or about the cutting already under way that nobody announced. The rungs being passed now were passed by indexing pay and pensions far below inflation.
A capacity figure fails in three ways at once: 1) it treats roubles as interchangeable, so a rouble saved on a road project and a rouble saved on pensions are the same object to the accountant and entirely different objects politically, and ordering the options by size produces a sequence the state will not follow; 2) it counts gross, since cutting household payments yields goods rather than money — compressed consumption narrows the VAT base and part of the saving comes back as forgone revenue; 3) it assumes a cut is an announced decision, which is exactly what the 2026 cuts are not.
Nothing in the budget records a decision to reduce pensions. The reduction is there all the same.
Six rungs. The size of the saving plays no part in the ordering at all.
| Lvl | Rung | Who pays | Which balance degrades |
|---|---|---|---|
| 1 | National projects | The future, which cannot answer back | The long-run productive base |
| 2 | Transfers to regions | The region that goes to borrow from a bank | Local debt, and the region’s ability to fund recruitment |
| 3 | Public sector pay | The household directly | Consumption, and through it the VAT base |
| 3a | Pay for the security services and equivalents — pre-break | The coercive apparatus | Loyalty and manning |
| 4 | Pensions — the break | The loyal base | Consumption, and political resource |
| 5 | Military payments — terminal, irreversible | The recruitment mechanism itself | Reproduction of manpower |
| 6 | Reverse move: new payments with an empty treasury | Everyone, later | Political risk accepted above fiscal risk |
Rungs 1 and 2 are passed. Rungs 3, 3a and 4 are partly engaged in concealed form, and partial passage does not move the zone, which is why the widget reads 1–2 of 6 while the text says rungs 3 and 4 are already being walked.
Rungs 2 and 3 are kept apart though both cut a transfer. Cutting transfers to regions moves the bill to a body that then goes to borrow: the bill is deferred, the payer stays silent, and the effect shows up in regional debt rather than in anyone’s reaction. Cutting wages moves the bill straight to the household, which reacts. Same accounting operation, different payer, different rung.
At a working deflator of 13–15%, the 2026 decisions read as follows.
| Group | Indexation | Prices outrun the payment by |
|---|---|---|
| Insurance pensions, from 1 Jan | +7.6% | 7.0% |
| Public employees under the ‘May decrees’ — teachers, doctors, social workers, culture — from 1 Jan | +7.6% | 7.0% |
| Social pensions, from 1 Apr | +6.8% | 7.8% |
| Military personnel, equivalents, other civil servants, from 1 Oct | +4% | 8.7–10.7% |
No decision to cut was taken and nobody protests against a raise = a rung passed with the payer silent. The result surfaces only in the VAT base and in retail turnover, thus a ladder built on announced decisions would show nothing here, while a ladder built on payers shows it immediately.
It looks as though the 7.6% group is favoured and the 4% group is not. That reading is wrong, and the construction had to be protected against it.
The 7.6% indexation is tied to the ‘May decrees’, under which pay for those categories is formulaically linked to the regional average wage. The formula is not discretionary: it requires growth in line with nominal wages, which is 13–15%. What was granted was 7.6%. That group is under-indexed as well, relative to a legal obligation rather than to inflation = one tier of underpayment measured against two different bases.
The ordering encodes the sensitivity of the payer and is not a forecast of sequence. A later rung can be engaged while an earlier one is not, and that is recorded as observed rather than smoothed away, since no other observation carries the same information.
The 2026 indexation calendar is that sequence written down: pensioners and the ‘May decree’ categories protected first and in January, social pensions in April at a worse rate, the security services, the military and the rest of the civil service last, in October, at 4%. The group with the strongest formal claim to protection was cut hardest and latest.
What follows from that is a hypothesis, and it is labelled as one. Part of a security officer’s compensation is non-monetary: service in the security structures confers de facto protection from contract recruitment and mobilisation pressure. Against an alternative of a signing bonus of ₽1.65 mn plus the risk of death, that protection has measurable value, and the regime most probably uses it to pay for the indexation gap = the last non-monetary asset has been monetised, and it is exemption from the war.
One series tests it: the share of interior ministry and national guard personnel leaving for a military contract, together with undermanning in regional units. If protection works as payment, the flow is low; a rising flow means the non-monetary compensation has depreciated and the gap has become uncovered. The series is not collected, and the slot is reserved rather than filled with an estimate.
Further cuts to national projects cost nothing today and the resource is nearly exhausted, having been cut three times. Transfers to regions run into regional debt at a fifteen-year record, and the consequence has already been observed, when regions cut contract signing bonuses in late 2025. Under-indexation yields goods rather than money. An open cut to the security services hits the manning of the very apparatus used to suppress reactions to everything else on the list. Military payments are terminal.
The date binds harder than the amount. The decision cannot be taken before October, so roughly a quarter of the annual capacity applies to what remains of 2026, and the part being cut is where payments on already executed contracts sit. A fourth-quarter sequestration of commitments already entered into is non-payment rather than saving: the bill passes to the supplier as accounts payable and returns as a line of defence-industry working capital. The marker is a rise in overdue budget payables together with a rise in bank working-capital lending to defence order contractors, in the same quarter as the announced cuts.
It needs no access to internal documents. Indexation rates, budget lines and legal obligations are all published, and the state publishes them because it wants credit for the raise.
Three failure modes.
A state that has to cut announces the cut it can defend. What it does quietly, through a formula and a percentage, is the part that names the payer.