On 8 October the Russian central bank published issue No. 10 (115) of its Monitoring of Sectoral Financial Flows, covering payments through the Bank of Russia's own payment system to 30 September. A month ago we read the same series for August and found incoming payments to state-demand industries 20.0% below a year earlier. September is 24.4% below, and with September the quarter closes: for the third quarter as a whole the figure is 22.6%. That is the third consecutive quarter in which this group of industries received less than a year before, and each has been deeper than the last.
The quarter, in the bank's table 10
Table 10 reports growth in incoming payments weighted by industries' shares of value added, without seasonal adjustment, against the average of the corresponding period of the previous year. Read down the row for state-demand industries and the slide is a shape, not a month:
| period | vs a year earlier |
|---|---|
| Q4 2025 | +1.2% |
| Q1 2026 | −2.2% |
| Q2 2026 | −14.0% |
| Q3 2026 | −22.6% |
| September 2026 | −24.4% |
The last reading in the series that was above water is the final quarter of 2025. Table 1, which measures the same group against the previous quarter's average with seasonality removed, gives September at −15.4% and the third quarter at −13.1%, and its own year-on-year column reproduces the −22.6%; the two tables cross-check.
Nothing else in the publication moves like this. Over the same September, payments to consumer-facing industries were 9.2% above a year earlier, to intermediate-demand industries 8.7% above, and to external-demand industries 15.6% above. Investment-demand industries were 8.3% below — bad, and improving from −13.7% in July. The total daily flow weighted by industry shares of GDP was 3.6% above a year earlier. The bank's summary opens on a different measure of the same month: payments 3.3% above the average level of the second quarter, seasonally adjusted.
One class is the whole of it
The group called state-demand industries is not one activity. Table 3 breaks it into classes of the Russian activity classifier with their weights inside the group, and the arithmetic is unambiguous. Against the group's September fall of 15.4% on the seasonally adjusted measure:
- class 84, “activity of bodies of state administration”, weight 60.4% of the group, contributes −17.0 percentage points;
- education, weight 18.1%, contributes +0.8;
- health care, weight 15.0%, +0.6;
- social services without accommodation, arts and sport, together 4.8% of the group, +0.4 between them;
- the four remaining classes take off 0.3 between them.
The contributions are the bank's own, each rounded to a tenth of a point, so the ten of them come to 15.5 rather than exactly to the group's 15.4.
One class subtracts more than the entire group falls. Everything that is not public administration — schools, hospitals, social services, culture, sport — is adding to the flow, slightly. The fall is in one line, and it is the line that is three-fifths of the group by weight.
That makes the second exclusion in table 10 worth stating exactly, because it is the one pair in the publication that differs by public administration and nothing else: the daily flow excluding extraction and petroleum products was 1.7% below a year earlier in September, and excluding extraction, petroleum products and public administration it was 2.5% above. The 4.2-point difference between those two is what public administration does to the aggregate.
What this series is, and what it is not
The bank's own methodological comments set three limits, and they decide what the number may be used for.
Coverage is partial. These are payments through the Bank of Russia's system, and “they cover less than half of the payments made, taking into account payments on LORO/NOSTRO accounts, intra-bank transfers and transfers through other payment systems”.
The data are winsorised: a boundary is set at “the 99.5 percentile of the distribution of non-zero payments” for each industry and payments above it are replaced by that boundary, so large one-off transfers do not enter at full size. This measures the breadth and regularity of payments more than their total volume — a reader who turns −24.4% into “state spending fell by a quarter” has read it wrong, and so would we if we wrote it.
The direction is inward. The series counts payments arriving in the accounts of these industries, sorted by the recipient's main activity code in the state register. For class 84 that is money landing in the accounts of state bodies, not money they send out to contractors. The bank does not split the class — not into federal and regional, not into defence and civil — so what fell inside it is not in the document, and we do not claim it.
Why we watch it anyway
Because of its position in the queue. Our own measure of unpaid bills between companies — Rosstat's overdue receivables — arrives a quarter late. This series arrives eight days after the month ends.
What it shows is one cluster, defined by who buys from it: the part of the economy whose customer is the state. Inside that cluster the class that fell is the state's own administrative bodies, and what fell is money arriving in their accounts. Whether less arriving means less leaving is not in the table, and this is where a month ago we stopped short of a claim we could not support. The length is what is new. A quarter is long enough that a single month's shifted payment date cannot explain it, and an order placed but paid late reaches the supplier exactly as an order not placed: a rouble not received.
The same autumn has made the other end of the chain more expensive. The bank's September risk review, published the same day, records that spreads to government bonds for the weakest investment-grade corporate borrowers widened by 407 basis points in a month, concentrated in construction and leasing. A supplier waiting to be paid and borrowing more dearly to wait is the mechanism our weakest indicator measures after the fact.
What this changes in our model
Nothing directly. A payment-flow series is not one of the eight components, and no analytical publication enters the index.
Overdue receivables remain the weakest reading we carry: 1.96 times their January 2022 level in real terms, on data to the end of July, scoring 4 of 100 against that pre-war anchor of 1× and a registered trouble line at 2×. The June reading was 1.91× and scored 9. The strength index stands at 46.2 of 100, in the stress zone; this morning's run moved no reading at all.
What we do not know
Whether this is a cut or a delay. The series records payments, not obligations. Money not paid in September may be paid in December; on this measure a deferral and a cut look the same. That is exactly why we read it as a leading signal for receivables rather than as a spending figure.
How much of the fall is the new weights. The base year for value-added shares moved from 2018 to 2023 in July 2026, inside the comparison we are reading, and the document does not quantify the effect. What the document does show is that the slide started before the change: the second quarter, computed on the old weights, was already 14.0% below a year earlier.
Why the bank does not discuss it. The publication's summary and its five bullet points cover external, consumer, investment and intermediate demand and the federal districts. The state-demand group, which holds the largest movement in the tables, is named in neither — a description of the document, not a motive, which we cannot read.