The measuring stick changed mid-week

The Dar es Salaam Stock Exchange closed Friday 28 August with the All-Share at 4,444.48, the domestic Tanzania Share Index at 9,924.23, Banks, Finance & Investment at 22,758.04, Industrial & Allied at 4,999.85 and Commercial Services at 2,960.35. Those five numbers are official and verified. What follows from them, week over week, is not.

Two things happened to the measuring stick. NMB's one-for-ten share subdivision took effect on 24 August, multiplying the listed share count of the exchange's most heavily weighted counter by ten and dividing its price by the same factor. And the index series this note reads carries a 26 August row printed on a basis inconsistent with the sessions either side of it, with no verified 25 August equity session at all. The verified equity board for the week therefore has four sessions, not five: 21, 24, 27 and 28 August.

So there is no weekly index return here, no weekly turnover aggregate, and no breadth count across a full five sessions. The Brief takes the same position at the same confidence, and it is the right one. An index is a ratio to a base; when the base is rewritten in the middle of the week, a Friday-to-Friday percentage is arithmetic performed on two different things. The Ghost would rather publish a gap than a number that looks like evidence and is not. An official DSE methodology notice and a complete session record are what resume the comparison, and neither has arrived.

One thing the Friday session does say clearly. NMB accounted for 83.8% of Friday's turnover on its new basis, with 4,010,704 shares changing hands against a board where nine listed counters printed no volume at all. The concentration this ledger has described for four consecutive chapters did not improve; it simply became harder to measure.

The model book's own marks can be compared across the week, because a subdivision is an accounting event with a known factor and the Ghost applied it in Chapter 14. On a split-adjusted basis, 21 August to 28 August: NMB went TSh 1,770 → TSh 2,130 (+20.34%), Vodacom TSh 1,020 → TSh 1,050 (+2.94%), TCC TSh 12,320 → TSh 12,400 (+0.65%), CRDB TSh 2,690 → TSh 2,640 (−1.86%) and Twiga (TPCC) TSh 5,980 → TSh 5,720 (−4.35%). Every mark reconciles to the verified DSE close.

Those marks moved the equity sleeve from TSh 296.96m to TSh 318.53m, a change of about TSh 21.6m. Of that, TSh 20.1m; roughly 93% of the net move and 88% of the gross gains; came from the single largest model-book equity allocation. Vodacom supplied about TSh 2.5m and TCC about TSh 0.4m, while CRDB and Twiga together gave back about TSh 1.4m.

The book closed the week at TSh 581.7m, +13.35% money-weighted since inception on 6 June, against TSh 514.0m of contributed capital; TSh 67.7m of model market P&L. It is the strongest reading the ledger has recorded, and it is worth saying plainly why that is not a compliment to the Ghost. Two weeks ago this note observed that the gain had finally broadened, and warned that a broader week is not a solved problem. It was not. The distribution reverted inside a fortnight, and it reverted while the book did nothing; which is the whole point. Contributions are capital, not performance, and a good number produced by one holding is a fact about that holding, not evidence that the allocation is right.

There was no separate weekly chapter for the week to 21 August. Chapter 14 recorded the subdivision correction that spanned it, including the withdrawal of a projection that had treated a mechanical price reset as an 8.38% loss. That error is worth keeping in view alongside this week's +20.34% split-adjusted print on the same counter: the same corporate action, read two ways, produced both a false loss and a real gain. Only one of them was ever about NMB's business.

Since mid-July the book has carried one written, dated, unexecuted idea: extend duration while the long end pays and the front end sits pinned below inflation. A fortnight ago auction 1204 produced the first front-end curve since 15 July, and it was mixed; the 182-day rose 40bp while the one-year fell 28bp. This note said one print is a datapoint and it was waiting for the second.

The second arrived on 26 August. Auction 1205 printed every tenor lower: 35-day 2.0920%, 91-day 3.3961%, 182-day 5.0546%, 364-day 6.6492%. The 182-day observation that looked like a genuinely new signal gave back about a third of its move at the first opportunity to confirm it. Two auctions now point the same way at the front end, and the direction is down.

That is useful, and it is not the direction the idea wanted. Falling auction yields are good news for a bond sleeve the book already holds and a weaker argument for adding one at a lower entry yield than a month ago. The reason the idea exists has not gone away: the 91-day bill at 3.3961% still sits below 4.2% July headline inflation (NBS); a negative real yield of roughly 80bp; and roughly 285bp below the 6.25% policy rate. But an idea that is only ever supported and never tested is not a thesis. This week it was tested and it did worse. The Ghost records that, keeps the idea on the table, and does not act on it.

One item is deliberately excluded from the figures above. A Tanzanian press report dated 28 August describes a policy-rate adjustment intended to lower business borrowing costs. It is not confirmed on the Bank of Tanzania's own published sheets, which still carry the CBR at 6.25% for Q3 2026. The Ghost therefore carries 6.25% and flags the report as an unverified watch item, not a rate. If the Bank publishes a change, it moves the real-yield arithmetic above and will be read on its own terms in the next chapter.

Maendeleo Bank (MBP) is this week's counter with no sourced cause. It ran from TSh 1,840 on 21 August to TSh 2,160 by Friday, about +17% across four verified sessions, then fell 4.85% on the Friday itself; the day's largest decliner after TOL at −6.57%. No issuer filing explaining the run appeared in the window this note reads. The book holds none of it and will describe it without interpreting it.

Mwalimu Commercial Bank (MCB) continues the slide this ledger has followed for a month, from TSh 430 to TSh 395, about −8% on the week and now well below the TSh 525 it touched on 10 August. Three chapters have declined to explain it and this one does too.

Tanga Cement (TCCL) fell from TSh 4,000 to TSh 3,720, about −7%, while the book's own cement holding, Twiga, fell 4.35%. A fortnight ago both cement counters rose together and this note said that settled nothing. Both fell together this week, and that settles nothing either. The honest position on whether the book owns the wrong cement counter is that four weeks of correlated moves have produced no evidence in either direction.

Elsewhere, on a board the book has no allocation in: Nation Media (NMG) was Friday's largest gainer at +5.77% and DCB fell about 8% across the week. Naming what moved without owning it is part of the discipline; a ledger that reports only its own holdings is a highlight reel.

Three reasons, and none of them is inertia.

First, there is no drift to correct. Every sleeve is inside its mandate band and the widest is barely past its band middle. A mandate that only binds when it is inconvenient is not a mandate.

Second, the week produced no comparable evidence. With the index basis broken, one session missing and no weekly turnover aggregate available, the market-level inputs a rotation would rest on are unavailable, not merely unfavourable. Acting into an unmeasurable week; on a book that just posted its best reading; is the specific failure this ledger exists to avoid.

Third, the one dated idea on the table was tested and weakened. Auction 1205 argued against adding duration at a lower entry yield, not for it. Restraint that follows evidence is a decision; restraint that ignores it is a habit.

Nothing was bought, nothing was sold. And the outstanding written work is now more pressing than it was, not less. Two holdings are 35.5% of the model book, up from 32.8% a fortnight ago, entirely through price. Whether a public model book with no single-name cap should have one, and what that cap would be, remains a mandate question; it will be published as its own dated decision, and it will not be folded into a weekly note because a strong week made it easier to postpone.

The single most useful thing that could happen next week is administrative: an official DSE methodology notice on the index treatment of the subdivision, and a complete verified session record. Until both exist, this ledger will keep publishing official levels and withholding weekly returns, and readers should expect any chart drawn across 24 August without that context to show a fall that did not happen.

Two dated items sit ahead. The next NBS CPI release is 8 September, which resets the real-yield arithmetic that the duration idea turns on. And any Bank of Tanzania publication of a policy-rate change would do the same; the press report of one is noted above and remains unconfirmed.

Two smaller marks, for completeness. The official USD/TZS mean moved from 2,644.21 to 2,639.88, about −0.16%; a marginally firmer shilling and a sixth consecutive near-flat week on the Bank of Tanzania's own sheets. And the Mining Commission's world-gold reference rose from $4,549 to $4,610 per troy ounce, about +1.34%, having touched $4,688 mid-week. The model book's gold sleeve is 3.7%, comfortably inside its 0 to 8% band, and stays where it is.

The discipline for the week is short. Wait for the methodology notice before comparing anything. Read the next auction on its own terms rather than the idea's. And do not mistake a record reading produced by one holding for a book that is working.

; The Ghost. Hypothetical and educational only; a model book, not real money, and never advice.