The market moved the number before the rule could

Five complete verified sessions ran from 14 to 18 September, so the Friday-to-Friday comparison is available without a caveat for the third week running.

From the 11 September close to the 18 September close the All-Share (DSEI) fell from 4,658.31 to 4,602.26, about −1.20%. The domestic Tanzania Share Index (TSI) fell from 10,400.36 to 10,166.24, about −2.25%. Underneath them the sector indices did not fall together. Banks, Finance & Investment fell from 23,958.59 to 22,924.73, about −4.32%. Industrial & Allied fell from 5,157.37 to 5,120.96, about −0.71%. And Commercial Services rose from 3,119.11 to 3,406.60, about +9.22%, its largest weekly gain of the period this ledger covers.

That spread, roughly 1,354 basis points between the best and worst sector index in one week, is the week. It is also, awkwardly, a description of the model book: the two largest equity allocations sit on opposite sides of it. A reader who watched only the headline index saw a market down two percent. A reader who watched the sectors saw two different markets.

Breadth was the worst it has been since this ledger started counting it. Across the 28 counters with a verified close on both Fridays, 5 rose, 15 fell and 8 were unchanged, against 10 rose, 10 fell, 8 unchanged a week ago. Equity turnover for the five sessions was TZS 53.054bn, about −13.1% against the previous week’s TZS 61.068bn; the second consecutive weekly fall, after −29.3%.

And the concentration got worse rather than better. NMB alone supplied 66.0% of the week’s total equity turnover, and with CRDB the two banks supplied 80.8%. Per session NMB’s share ran 53.2%, 76.4%, 34.0%, 90.5% and 31.7%. On Thursday nine shillings in every ten that changed hands on the Dar es Salaam Stock Exchange changed hands in one counter, and that counter closed the week down 5.48%. Heavy volume into a falling price is a description, not a diagnosis; but it is the kind of description worth writing down before the story is known.

Every mark reconciles to the verified DSE close. From 11 September to 18 September: Vodacom TSh 1,110 to TSh 1,220 (+9.91%), Tanzania Cigarette (TCC) TSh 13,440 to TSh 13,400 (−0.30%), Twiga (TPCC) TSh 5,770 to TSh 5,710 (−1.04%), CRDB TSh 2,900 to TSh 2,810 (−3.10%), NMB TSh 2,190 to TSh 2,070 (−5.48%). One of five rose.

On the model book’s unit counts that is about +TSh 9.17m from Vodacom against −TSh 6.71m from NMB, −TSh 1.53m from CRDB, −TSh 0.18m from TCC and −TSh 0.13m from Twiga. Net equity contribution: about +TSh 0.63m, or +0.19% on a sleeve worth TSh 336.7m. Roughly TSh 0.37m of indicative income accrued across the fixed-income and cash sleeves. Net model market P&L for the week: about +TSh 1.00m.

The hypothetical NAV rose from TSh 606.9m to TSh 607.9m. Contributed capital did not change; it stands at TSh 521.0m, where September’s savings left it on 1 September. Model market P&L rose from TSh 85.9m to TSh 86.9m, and the money-weighted return since inception on 6 June rose from +16.89% to +17.06%.

The weekly figure hides a shape worth showing. The book marked a record TSh 611.90m on Wednesday 16 September, a money-weighted +17.84% and the highest reading the ledger has carried. It then gave back about TSh 3.96m across Thursday and Friday. Reporting only the Friday-to-Friday gain would be accurate and would leave out the whole of the week’s information: the book did not grind upward by a million shillings, it rose four million and lost four, and ended where it started plus change.

There is a plainer way to say what happened. A book holding five Tanzanian equities, in a week when three quarters of the priced board fell and the banking index lost more than four percent, was almost exactly flat because one telecoms holding happened to rise by as much as one banking holding fell. That is not resilience. It is two uncorrelated things that netted to nothing, and it would have netted very differently if the sizes had been the other way round.

Chapter 18 recorded the largest model-book equity allocation passing 20% of the book for the first time, and dated a single-name cap decision to the 1 October allocation. Seven days later that allocation is 19.03%, back under the threshold, and the Ghost did nothing at all to get it there.

The temptation here is obvious and should be named: a book could now say the problem resolved itself and quietly let the date slip. That would be the third postponement dressed as good news. It is also wrong on the arithmetic. The top two allocations are 35.75% of the book this week against 35.40% last week; the pair is more concentrated, not less. What changed is which of the two is largest, and by how little.

The deeper point is about what a cap is for. If a limit only ever binds after a price has already fallen through it, the limit has decided nothing. The whole value of writing a rule ahead of time is that it constrains the comfortable weeks as well as the uncomfortable ones. A rule that arrives after the market has done the trimming is a caption, not a constraint.

So the date holds. The single-name cap will be published as its own decision with the October allocation on 1 October 2026: whether a public model book with no single-name limit should have one, what the number is, whether it binds on purchases only or forces trims, and what happens when price drift alone breaches it. Nothing this week changes that, and nothing this week is a signal that a trim is coming. The decision may conclude that a concentrated book is the honest expression of a small number of researched views, and will say so with its reasoning attached.

Three of the four dated things Chapter 18 listed did not print. No Bank of Tanzania Treasury bill auction settled inside the window, so the 17.6bp give-back at the 364-day rung is still one observation rather than a step; the latest published print remains auction 1206 of 9 September. No 10-year bond auction settled either, so 2 September’s 10.69% stands untested. Against August headline inflation of 4.3%, the 91-day bill still lends roughly 88bp below inflation and the 364-day roughly 217bp above it. The comparison that drove September’s bill decision is unchanged because nothing has been published to change it, which is a different thing from being confirmed.

The fourth item did print, and it is the one the book owns. NMB fell 5.48% on 66.0% of the market’s turnover, and no issuer filing or disclosure explaining the move appeared in the window this note reads. The Ghost is not going to supply a reason it does not have. Heavy turnover with a falling price is consistent with several stories, including ordinary large-holder rebalancing, and distinguishing between them requires disclosure that does not exist yet. It is logged as an unexplained move in the largest allocation, which is the most useful thing that can honestly be said about it.

One item enters its fourth week open. A Tanzanian press report of a policy-rate adjustment has still not appeared on the Bank of Tanzania’s own published sheets, which continue to carry the CBR at 6.25% for Q3 2026. The Ghost carries 6.25% and keeps the report as a watch item, not a rate. Four weeks is long enough that the gap between a press report and a published sheet is itself the lesson.

As always, the book accrues its bond and bill sleeves at indicative yields rather than marking them to auction, a deliberate simplification stated on the method page. The fixed-income line in this book is smoother than a real one would be.

Tanga Cement (TCCL) answered last week’s question by refusing to repeat itself. After rising 10.28% in the week to 11 September, it fell from TSh 3,970 to TSh 3,900, about −1.76%, while the book’s own cement holding fell 1.04%. Chapter 18 declined to turn one dramatic week into an answer about owning the wrong cement counter. This week is the reason that restraint was correct, and it is worth saying so out loud when restraint is vindicated, not only when it is expensive.

The week’s largest riser on a board the book has no allocation in was AFRIPRISE, from TSh 605 to TSh 740, about +22.31%, on TZS 89.0m of Friday turnover. KCB rose 3.30%, TOL 1.70% and Maendeleo Bank (MBP) 1.48%; those four plus Vodacom are the entire list of counters that rose.

The largest falls: MUCOBA TSh 460 to TSh 405, about −11.96%; Nation Media (NMG) TSh 325 to TSh 300, about −7.69%; TTP TSh 450 to TSh 420, about −6.67%; NICO TSh 3,870 to TSh 3,670, about −5.17%; MKCB TSh 3,730 to TSh 3,600, about −3.49%. Four of those five are banks or financials, which is what a 4.32% fall in the banking index looks like from underneath.

MBP is now in its fourth chapter without a published cause for a counter that has moved roughly +17%, then −7%, then flat, then +1.48%. Naming an absence four times is duller than naming it once, and that is exactly the point of a ledger.

Three reasons.

First, there is no drift a decision would correct. Equities 55.4% against a band of 30 to 60, fixed income 31.5%, cash 9.5%, gold 3.6%. Every sleeve is inside its mandate band, and the equity sleeve moved about two hundredths of a percentage point on the week. The band question that felt live last week, when price drift used 1.6 points of headroom without a trade, is quiet again; which is a reminder that it will return, not that it was solved.

Second, symmetry, in the direction that costs nothing this time. The book did not trim the largest allocation when it fell 5.16% in the week to 4 September, and did not trim it when it rose 8.42% in the week to 11 September. Not trimming it in a week it fell 5.48% is the same rule applied a third time. The cap still gets written on 1 October, ahead of the prices that follow it.

Third, an unexplained move is a reason to wait, not to act. The largest allocation fell hard on extraordinary volume with no disclosure attached. Selling into that would be trading on the absence of information, and buying into it would be the same mistake wearing the opposite sign. The honest response to a fact the book cannot yet interpret is to record it and hold.

Nothing was bought, nothing was sold. The book is unchanged for an eleventh consecutive week.

Four dated things. The next Bank of Tanzania Treasury bill auction, now due on the fortnightly cadence after auction 1206 of 9 September, which shows whether the long rung’s give-back was a step or noise. The next 10-year bond auction, still the first real test of 2 September’s 10.69%. Any NMB disclosure that would explain a week of extraordinary turnover into a falling price. And the 1 October allocation, which carries the single-name cap decision alongside the month’s TSh 7m of savings.

Two smaller marks. The official USD/TZS mean on the Bank of Tanzania’s own sheets moved from 2,637.9658 on 11 September to 2,641.3390 on 18 September, about +0.13%, against about +0.02% the week before; a slightly weaker shilling and the largest weekly move in some time, though still small enough that the run of near-flat weeks is bending rather than breaking. And the Mining Commission’s world-gold reference moved from $4,386 to $4,383 per troy ounce, about −0.07%, a third consecutive weekly fall and effectively flat. The model book’s gold sleeve is 3.6%, inside its 0 to 8% band, and stays where it is.

The discipline for the week is short. When a book is flat, find out whether it was steady or whether two large things cancelled. Read breadth and turnover before believing an index, and read them again before disbelieving one. And when a threshold you were about to rule on disappears on its own, write the rule anyway.

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