Where you are. Lesson 11 mapped where idle cash sleeps, with the T-bill in the best corner setting the benchmark, and lesson 12 opened up the repo: a secured overnight loan whose lender keeps the bond if the borrower fails. What you hold is a map of parking spots ranked by safety and speed back to cash. What you have not yet done is stare at the word safety itself. This lesson walks the same pile asking what happens when the borrower behind each spot fails, then meets the product built to spread that danger across many instruments and many holders - and works the arithmetic showing that spreading a loss is not the same as escaping it.
The number that never moves
Watch a treasurer’s screen for a morning. Every price on it breathes. Bonds drift, currencies tick, and even the T-bill you priced in lesson 11 crept up overnight, the way a discount instrument does as its maturity shortens. Except one row. One row prints 1.00, printed 1.00 yesterday, and has printed 1.00 every trading day for years: the share price of a pooled fund where thousands of companies park their cash. It looks like the safest number in finance. It is certainly the stillest.
This lesson asks what that 1.00 is standing on. Then it breaks one of the legs, and watches the stillest number in finance move.
The idea in one paragraph
The cash pile carries credit risk - the risk that the somebody behind a claim fails and the claim pays back less than promised. Lesson 11’s parking spots are a gradient of it: the T-bill is the government’s own promise, the repo lender holds collateral, but a bank deposit above the insurance cap and a company’s commercial paper are unsecured promises, and unsecured promises can lose. A money market fund is the market’s product answer: pool many holders’ cash, spread it across the whole gradient, and hold the share price at exactly 1.00. The spreading is real - a failure that would have cost one holder heavily costs every holder a little - but it is only spreading. When one slice of the pile loses, the loss flows straight through to the share price and lands on every holder in proportion to what they hold. In this lesson’s stylised numbers: a 10% loss on a 20% slice, and the stillest number in finance prints 0.98.
Rank the pile by what stands behind it
Run down lesson 11’s parking spots asking one question of each: if the borrower behind this claim failed tomorrow, what would stand between the pile and the loss?
The T-bill. The government’s own short-dated IOU. In its own currency the government is the strongest borrower on the map, which is exactly why lesson 11 made the T-bill the benchmark everything else is measured against. It is the pile’s definition of safe - a definition, note, not a law of physics.
The repo loan. Lesson 12 already answered this one in code: the borrower fails and the lender keeps the bond. The credit question does not vanish; it changes subject, from “will they pay?” to “is the collateral good?”. Secured is the word for that trade.
The bank deposit. Below the insurance cap, an insurer stands behind it. Above the cap, nothing does - and you already know the mechanics from lesson 4. A bank’s assets must equal its liabilities plus equity, so when the assets rot, equity absorbs the loss first; once equity is gone, what remains lands on the unsecured claims, the above-cap deposits among them. Not villainy: the balance-sheet invariant you have been asserting since lesson 3, running in the direction nobody enjoys.
Commercial paper. The corporate IOU from lesson 11’s map: a large company borrowing for weeks or months, unsecured, on nothing but its name. If the company fails, the paper joins the queue of unsecured creditors. This is the risky end of the pile, and the extra yield it pays over the T-bill is not generosity; it is the price of standing in that queue.
Spread the parking
Now be the treasurer with a pile far too big for the insurance cap. Park it all at one bank and everything above the cap is a single unsecured bet on a single name. Split it across a handful of banks and you hold several smaller bets and a lot of accounts to reconcile. Buy nothing but T-bills and you give up the yield the unsecured instruments pay for being losable. The market’s answer is to pool the problem:
Place it on the two-tier map before going further, because the fund is a fresh instance of a shape you already know. Your share is a claim on the fund; the fund’s assets are claims on the government, on banks, on companies. Module 0’s map lesson chased an e-money balance through exactly this structure - claims on claims, one more link between you and the floor. Nothing sinister. But every link is a place where a promise can fail, and the fund’s links are the pile itself.
Inside the fund, the pile is run in slices: a T-bill slice, a repo slice, a deposit slice, a paper slice. Fund managers call each slice a sleeve, and the word is worth adopting because it names the unit at which credit risk arrives. A failure never hits “the fund”; it hits a sleeve. The question is what happens next.
Push a loss through the sleeves
Build the fund with stylised numbers - round, chosen for clean arithmetic, taken from no real fund. Per 1.00 of shares it holds 0.40 in T-bills, 0.25 in repo lending, 0.20 in bank deposits, 0.15 in commercial paper. The sleeves sum to exactly 1.00, and that is not a coincidence; it is the definition. The share price is nothing but the sum of the sleeve values, per share.
Now fail a bank. The deposit sleeve sits above the cap at a bank that goes down, and the fund expects to recover all but 10% of that sleeve: a 10% haircut, reusing lesson 12’s word for a slice shaved off a value, this time shaved off a claim rather than off collateral. The sleeve marks down from 0.20 to 0.18, and the share price becomes 0.40 + 0.25 + 0.18 + 0.15 = 0.98. The drop is the sleeve’s weight times the sleeve’s loss: 0.20 × 0.10 = 0.02. In words: the fund scales the loss down by the sleeve’s share of the pile, then passes every remaining unit of it straight through to the price.
Through to whom? Everyone. The fund has no mechanism for choosing which holders absorb a loss: every share is an identical claim on the same pool, so when the pool is worth 2% less, every share is worth 2% less. The phrase is pro rata - in proportion to holdings - and it is the product’s whole social contract in two words.
Count what the spreading bought, in both directions. The treasurer whose entire pile had sat at the failed bank would be staring at the full 10%. Inside the fund, the same failure cost 2% - five times smaller. But the treasurer who had never heard of that bank also paid the 2%. The fund resized the loss and redistributed it; at no point did it destroy any of it. The sleeve lost 0.02 per share, and 0.02 per share arrived, in full, on every holder’s screen. Spreading is not escaping; it is a choice about the shape of the arrival.
Check yourself
1. A repo loan and an above-cap deposit both depend on a bank that can fail. Why does the deposit sit further down the safety gradient?
Because of what stands behind each promise. The repo lender holds the bond: a borrower failure becomes a question about the collateral’s value, not a loss of the principal. The above-cap depositor holds nothing but the bank’s IOU: when the bank fails, equity absorbs losses first, and whatever remains lands on the unsecured claims, the deposit among them. Same borrower, different backing - and the backing decides the rank.
2. The fund held four sleeves and only one was hit. Why does every holder lose, and why exactly 0.02 per share?
The share price is the sum of the sleeve values per share, and every share is an identical claim on the same pool, so the fund has no way to assign a loss to some holders and not others; it can only pass it through pro rata. The deposit sleeve was 0.20 of each share and lost 10% of itself, so each share lost 0.20 × 0.10 = 0.02: weight times haircut, and the price steps from 1.00 to 0.98.
3. A treasurer complains: “alone I only lose if my bank fails; in the fund I pay for every bank failure. The fund made things worse.” What is right and wrong in that?
Right: the fund converts a loss the treasurer might have dodged entirely into one they certainly share; spreading means paying a slice of other people’s failures. Wrong: the alternative was a concentrated bet that a self-chosen bank is the safe one, with the full 10% - not 2% - as the price of being wrong, and no reliable way to know in advance. The fund does not lower the pile’s total credit risk; it trades a rare ruinous loss for a small certain-to-be-shared one. Which shape is better depends on which loss you can survive.
4. Why is 0.98 a run trigger rather than a 2% inconvenience?
Because the product’s promise is exactly 1.00. Holders treat the shares as cash - parked, spendable, not an investment - and a print below one is proof the promise can break, not a market wobble. The rational response to breakable cash is to redeem before the next sleeve fails, and since that is every holder’s rational response at once, the 2% loss matters far less than what it announces. The failure is the broken promise; the number just carries the news.
Do this
Ten minutes, standard library only, working from module-01-money-at-rest. Open code/cash_pile_risk.py: the stylised four-sleeve allocation from this lesson, a share_price function that sums the sleeves, and one TODO(you) marker in apply_haircut, which must return a new dict with the named sleeve reduced by loss (a fraction of that sleeve) and every other sleeve untouched.
python3 code/cash_pile_risk.pyRun unmodified, the starter dies on NotImplementedError. Completed, both assertions pass and it prints two lines, ending with exactly
the fund spread the parking; the loss still reached every holderafter reporting share price 0.9800.
Then move the loss. Point the haircut at "commercial paper" instead of "bank deposits" and rerun: the script dies on its pinned assertion, and the number in the error message - 0.985, give or take floating-point dust - is the lesson’s arithmetic answering back, because a 0.15 sleeve times a 10% haircut is only 0.015 off the dollar. Module 0’s project preached property tests over example tests; this assert pins an example on purpose, one scenario at one exact price, which is why moving the loss trips it and why the error it raises still teaches. Restore the line before moving on. The completed version is in solutions/cash_pile_risk.py.
What you can now do. You can rank any cash pile by asking what stands behind each claim, and name the two spots where the answer is “nothing”: the above-cap deposit and commercial paper. You can push a loss through a fund’s sleeves by hand - weight times haircut, passed through pro rata - and state exactly what a money market fund’s spreading buys: a smaller loss for everyone in place of a ruinous loss for someone, never no loss at all. And you can say why 0.98 on a product that promises 1.00 is a failure, not a fluctuation. The next lesson gives this judgement its job title: every bank and large company runs a treasury desk that decides where the cash sleeps, trading safety against liquidity against yield - and the newest people running that desk, under a new name, are the stablecoin issuers module 6 puts under the microscope.