What may enter a financing identity

The arithmetic is checkable, thus the errors that survive are of a different kind: a line counted as a source when realising it produces the very thing the identity exists to constrain, or a line counted as reducible when it cannot be reduced. Three tests, each derived from a case that failed one, are applied before a line enters.

The sovereign wealth fund is the standard answer to how the gap gets closed, and it appears in the coverage line at 7% of the need for January–July. It is not treated as a source, for a mechanical reason.

The distinction is invisible in the accounts. Both routes show the same reduction in the fund and the same roubles arriving in the treasury, and only the counterparty differs — with it, whether new money was created. An identity that counts the fund as a source without asking who buys the asset will show the gap closing while the quantity the panel measures grows by the same amount.

A second, narrower point sits underneath: in 2026 the fund is not being drawn on at all, since hydrocarbon revenue has been above the base level for a fourth month and the fiscal rule blocks withdrawal. A source blocked by a rule the state set itself is not available merely because the balance exists. Both facts are stated, and neither alone would be sufficient.

The same test caught a larger error, in a construction the panel had published and has since withdrawn in full. Capacity to absorb the state’s claim had been estimated as the growth of the money supply, less the outflow into cash, plus the growth of bank liabilities to the central bank, yielding a headline conclusion that the state was taking around two thirds of the growth in banks’ resource base.

The money supply is the result of lending and budget spending, not a constraint preceding them: the more the state borrows and spends, the faster it grows = the share was measuring itself. Central bank data make the point directly — claims on the economy contributed 11.1 pp of the 12.9 pp annual growth in broad money. Deposits do not fund credit; credit creates deposits.

What replaced it is narrower and survives the test. Cash leakage exceeded ₽2.5 trn over the half-year while bank liabilities to the central bank rose ₽2.3 trn over the same period, thus the central bank is supplying roughly a rouble for every rouble of cash that has left the system = its refinancing is not additional capacity, it is replacement of a leak. Both quantities are published monthly, neither is constructed, and neither is produced by the other.

The withdrawal cost a conclusion the panel liked. What it bought is a result running against the earlier reading, now stated as the leading one: there is no volume constraint. The banking system can service a claim of around ₽16 trn, and the constraint lies in the price of the rouble and in who pays it.

Expenditure divides into a part that varies with tempo — munitions, equipment, replacement of losses — and a part close to fixed: pay for the standing force, payments to the families of the dead as an accumulated obligation, debt service, upkeep of industry. Cutting the tempo cuts the first group and barely touches the second.

Accumulated obligations do not compress because a payment triggered by an event that has already occurred does not respond to a decision taken afterwards. It is a stock being amortised, not a flow being purchased, and any identity modelling it as a flow will show savings that cannot be realised.

The observation anchoring this is put in the widget in its unflattering form: by results, Russia is already fighting no better than it would at reduced intensity, and lower intensity would have produced better results for Ukraine. The costs have not fallen.

• The borrowing plan is the finance ministry’s own obligation, and the shortfall is measured against it. Both sides of the quantity are published by the same body, so the measurement requires no estimate and no assumption about what the ministry privately expects. Reading the plan as a forecast, then generously observing that forecasts are often missed, discards the only hard number in the picture.
• Placements that do not happen create no coupon obligations, so the debt service ratio improves precisely because the market is closed. Alone, that line refutes the panel’s central numerical argument using the panel’s own data; paired with the placement figure, it is a reading.
• Sequestration capacity of ₽1.5–2.5 trn annualised is a 2027 quantity presented next to a 2026 gap. The in-year effect is ₽0.4–0.6 trn, under 5% of the need, since the decision cannot be taken before October.

The asymmetry disclosure sits beside the identity rather than inside it. The widget carries two columns — pressures working toward a wider gap and toward a narrower one — and the second is not perfunctory: inflationary erosion of accumulated rouble debt, five key-rate cuts reducing the cost of servicing floating-rate debt and subsidised lending, plants shut down by strikes not receiving damper payments, a closed market creating no new coupon obligations. Several of those weaken the panel’s own reading and are listed with the same specificity as the pressures that strengthen it.