Every month the Federation Account Allocation Committee distributes revenue to the federal government, the 36 states, the FCT and the 774 local governments, and every month the number is reported as though it were news. Some of it is. A large share of it is the calendar.
We indexed 123 consecutive months of state-level FAAC receipts — every state, every month, January 2016 to March 2026, sourced to the communiqués — against trend. The pattern is real: July is the month states receive the most, at 9.3% above trend, and it has run above trend in 8 of the last 10 years. March is the leanest, 6.4% below. The full swing from leanest to fullest month is 15.7 index points.
One thing to fix in your head before reading the table, because it trips almost everyone. FAAC runs on two clocks, a month apart. A communiqué published in August reports July revenue — the month the money accrued to the Federation Account. States receive that cash in August. The index below is on the receipt clock, deliberately: for a treasury that is 82% dependent on transfers, what matters is when the cash lands, not which month it accrued. So do not read the index month as the month a communiqué names. Translate: the communiqué due around 10 August 2026 reports July revenue, which states receive in August — an index of 106.8, not July’s 109.3.
That matters for a specific reason. The median Nigerian state gets 82% of its revenue from the Federation Account rather than its own tax effort. For a treasury that dependent, the shape of the year is not a curiosity — it is the working-capital constraint.
The seasonal index, month by month
100 = on trend. The index is a classical multiplicative decomposition: each month’s total state-pool receipts divided by the centred 12-month moving average around it, then averaged by calendar month across the series.
| Month cash is received | Detrended index | Observations below trend | Naive within-year index |
|---|---|---|---|
| January | 102.7 | 5 of 9 | 93.7 |
| February | 97.5 | 6 of 9 | 89.2 |
| March | 93.6 — weakest | 6 of 9 | 86.3 |
| April | 98.4 | 5 of 9 | 91.6 |
| May | 98.3 | 4 of 9 | 93.0 |
| June | 98.7 | 5 of 9 | 94.8 |
| July | 109.3 — strongest | 2 of 10 | 109.3 |
| August | 106.8 | 2 of 10 | 108.6 |
| September | 107.7 | 3 of 10 | 111.4 |
| October | 98.7 | 4 of 9 | 105.4 |
| November | 96.8 | 5 of 9 | 104.8 |
| December | 103.3 | 3 of 9 | 111.8 |
July through September is the reliable strong stretch. Q1 is the soft patch. On the detrended basis the second half of the year runs 5.9% above the first.
One honest limit before anyone builds a cash-flow model on this: the standard deviations run 7 to 13 index points. A single year can deviate substantially from the pattern. This is a tendency to plan around, not a schedule to bank on, and the difference matters if you are the one signing the contractor payments.
Two clocks, and why the difference is a month of cash
FAAC runs on two calendars at once, and Nigerian reporting uses both without always saying which. A communiqué issued in August reports the revenue that accrued in July. But the cash from that communiqué reaches state accounts in August. So “the July allocation” means July revenue to one writer and an August bank credit to another, and the two are a month apart.
The index on this page is on the receipt clock — the month the money lands. That is a deliberate choice, not a default. For a treasury drawing 82% of its revenue from the Federation Account, the question is when cash is available to meet payroll and certificates, not which month it accrued in. A seasonality index on the accrual clock answers a question no treasurer is asking.
The practical consequence: the strongest receipt month, July at 109.3, is carrying June revenue. When the August communiqué lands reporting a July revenue figure, that money is an August receipt, and August is 106.8. Read the two clocks as one and every seasonal claim you make is off by a month — which is why the tables in the Monitor label the national tier by revenue month and the state panel by receipt month, separately and on every page.
If you want the same index on the accrual clock, shift every row back one month: the strongest revenue month becomes June and the leanest February, both at the same values, because relabelling cannot change the sequence of ratios. One figure does move, and it is worth knowing which: the H2-over-H1 premium is +5.9% on receipts but +3.4% on revenue, because shifting the boundary moves a strong month out of the second half and a lean one in. Quote whichever matches the clock your reader is using, and say which one it is.
The number we nearly published instead — and why we didn’t
The right-hand column above is the version we built first, and it is wrong. It divides each month by its own calendar year’s average, which is the obvious construction and the one most likely to be reached for. It is contaminated by trend: nominal allocations grow through the year, so early months are pushed down and late months pushed up mechanically, by growth rather than by season.
The two constructions disagree materially:
| Claim | Naive within-year index | Detrended (correct) |
|---|---|---|
| Weakest-to-strongest swing | 25.5 index points | 15.7 index points |
| December | 111.8 — near the top of the year | 103.3 — mid-pack |
| H2 versus H1 | +19.2% | +5.9% |
| March below its benchmark | 10 of 10 years | 6 of 9 observations |
The naive version is more dramatic in every respect, which is exactly why it should be distrusted. A “10 out of 10 years” record is the kind of finding that gets quoted, and it dissolves under detrending.
One further test settles which is right. On the naive basis, H2 receipts exceed H1 receipts in 10 of 10 years — apparently overwhelming. But compare H2 of one year against H1 of the following year, where trend now works against the seasonal claim rather than for it, and H2 wins only 4 of 9 times. A genuine seasonal effect would survive that test better. Most of the raw H2–H1 gap was growth wearing a seasonal label.
Why we are publishing our own error
Because the alternative is asking you to take the rigour on trust. Every research shop on earth asserts that its numbers are careful. Very few show you the version that didn’t survive, and the arithmetic that killed it.
The ONYX position is that a figure is only worth something if you can see its source, its date, how it was captured, and how much to trust it — and that includes the figures we produce ourselves. A platform that will not publish its own correction has no standing to label anyone else’s data low-confidence.
What this is built on, and what it is not
The panel is 4,393 observations across 37 geographies and 123 consecutive months, all at high confidence, all fetched and hashed from RMAFC’s state allocation data, obtained via BudgIT. That series is the reason this analysis is possible: the seasonal index needs a decade of monthly history at state resolution, and reconstructing it from scratch is roughly two weeks of analyst time.
Three things are deliberately absent, and it is worth saying so plainly:
No local-government fiscal claim. ONYX holds a modelled LGA allocation series, apportioned down from the state figure using the RMAFC formula. It is labelled low-confidence and derived, because any variation in it is an artefact of the apportionment weights rather than an observation. We do not make distributional claims from it, and neither should anyone else.
No forward estimate of individual state receipts from the national headline. This one needs a second correction, because our first reason for refusing it was also wrong. We had said the state pool moves as a share of the national total across a range of 0.325 to 0.414 — too loose to split. That range was measured on the mismatched clocks described above. Paired correctly, with each month’s state receipts set against the previous month’s national distributable total, the eight most recent pairs (revenue months July 2025 to February 2026) give 0.349 to 0.392, roughly twice as tight. The original objection was mostly measuring our own alignment error.
The refusal survives on better evidence. Backtested out of sample — predict each month’s pool from the previous month’s national figure and the trailing 12-month ratio — the pool total comes within 2.9% on average, worst case 8.1%, across the twelve receipt months April 2025 to March 2026. That is respectable. But the pool is not the product; the individual state is. And state shares are stable only for the states nobody asks about. Over the last 24 months the median state’s share of the pool varies by 5.7%, while Lagos varies by 21.4%, Rivers 14.5%, Anambra 28.9% and the FCT 55.6% — because derivation and VAT make exactly those receipts lumpy. Compounded with the pool error, a forward-implied Lagos number carries about 22% error. A figure that wrong about the state everyone checks first is not a data product, so we do not publish one.
National figures are FAAC communiqué totals for the three tiers and the 13% derivation fund. The one exception is January 2026 revenue, for which no communiqué gave a figure. That month comes from the Accountant-General’s February 2026 disbursement report, Table I: ₦2.000trn, including a ₦100bn augmentation. State receipts are RMAFC data via BudgIT.
No interpolation. Our national series is built from FAAC communiqué headlines, and no communiqué ever announced a distributable total for January 2026 revenue. That is a gap in what was announced, not in what was paid. The January meeting stalled when state commissioners rejected the ₦1.969trn proposed for December 2025 revenue; that was resolved and paid at the start of February. January revenue was then shared at the February meeting without a headline release. The Accountant-General’s disbursement report later published it: ₦2.000trn to the three tiers and the 13% derivation fund, from ₦3.040trn gross, including a ₦100bn augmentation. The March communiqué itself uses January as its comparison month. We do not fill the national cell by interpolation; where it is filled, it is filled from the OAGF table and labelled as such.
The lag, stated rather than apologised for
The national tier totals publish monthly, roughly on the 15th of the following month. The state-by-state split publishes about a quarter later — that lag belongs to the whole ecosystem, not to any one publisher. Our state panel currently runs to March 2026, which is level with the public frontier, and the seasonal index above is computed from the full history rather than the newest quarter, so it does not degrade while the split catches up.
Where this goes next
This analysis is the analytical spine of the ONYX FAAC State Fiscal Monitor, which ships within 48 hours of each FAAC communiqué: the national number decoded against its seasonal norm, the full 37-row state panel with per-capita ranks and FAAC-to-IGR dependence, month-on-month movers screened for base effects, and the seasonal index recomputed. To see this read for two states you care about, request a free state note.
If you think the decomposition is wrong, we would rather hear it than not — [email protected]. The monthly pool history is published in the Monitor spreadsheet specifically so the index can be reproduced and challenged.