Everything else in the module is written in the language of one owner deciding how much of one business to hold. An institution running the same money writes the same ideas differently, because it has to answer a board, a regulator and an auditor, and because the person taking the risk is not the person measuring it. That vocabulary is worth learning properly. It is a second dialect for ideas you already hold rather than a correction of them, in the same way that AA1.01 and AA1.02 present mean-variance optimisation as the dialect of a multi-asset policy mandate. Risk here still means the probability and magnitude of permanent impairment. What follows is the apparatus that gets built around that idea when the capital belongs to somebody else.
How is risk governed in an institution?
Start with governance, because the sequencing is the point. Risk tolerance is decided at the top, in writing, before any position exists. A board or a set of trustees states how much loss the institution can bear and is willing to bear. Management converts that statement into a risk budget, which is the total tolerance sliced across strategies, desks or asset classes. An independent risk function measures the exposure against those limits and reports to the board rather than to the desk, and the front office takes risk inside them. The independence is the whole design. A measurement produced by the person whose bonus depends on the answer is not a measurement, and every large risk failure of the last four decades contains a version of that sentence.
Risk tolerance itself has two components and they are not the same thing. Ability to bear loss is a fact about the balance sheet: the size of the liabilities, when they fall due, the horizon, the liquidity that must be there on a date. Willingness is a preference, and it is the softer of the two. Where the two disagree the lower one binds, and the honest conversation is about educating the willingness rather than talking the ability upward. Your individual version of all of this already exists. Your investment policy statement is the board minute, the anti-ruin rules are the limits, and your decision journal is the risk report. The date to set the limits is before the drawdown. The date you will want to renegotiate them is during it, which is exactly why they are written down first.
Advertisement
What is the risk taxonomy?
Then the taxonomy, which is a checklist against a failure of imagination. The industry splits risks into financial and non-financial. The financial ones are market risk, meaning the risk that a price moves against you and which subdivides into equity, interest-rate, currency and commodity risk; credit risk, the risk that a counterparty or an issuer does not pay; and liquidity risk, which is really two risks wearing one name. Funding liquidity risk is the risk that you cannot finance your positions, and it is the one that killed LTCM. Market liquidity risk is the risk that you cannot exit at a price near the last printed one, and it is the one that turns a 4% position in a thin mid-cap into a 12% loss on the way out.
| Family | Risk | The question it asks | Where it has already appeared |
|---|---|---|---|
| Financial | Market | Does the price move against me, and how far | the volatility definition |
| Financial | Credit | Does the counterparty or issuer pay | E11.01, the credit workup |
| Financial | Liquidity (funding) | Can I keep financing what I hold | the forced seller |
| Financial | Liquidity (market) | Can I get out near the last price | the sizing cap |
| Non-financial | Operational | Does a process, a person or a system fail | M2.06, the control environment |
| Non-financial | Model | Is the formula answering a question the world is not asking | risk against uncertainty |
| Non-financial | Settlement | Does one leg pay while the other does not | MS1.03, payment and settlement |
| Non-financial | Legal and regulatory | Does the contract or the rule change under me | E11.04, governance forensics |
| Non-financial | Tail | Does a joint move nobody modelled arrive at once | hidden correlation |
Two of the non-financial rows are the ones individuals almost never name, and they are the two this module has been circling. Model risk is sizing under uncertainty in institutional clothing: a precise number produced from guessed inputs, trusted because it is precise. Tail risk is hidden correlation plus the fat tails the Quantitative lens puts a figure on. Naming them is not bureaucracy. A risk with no name has no owner, no limit and no report, and the first time anyone thinks about it is the morning it arrives.
What are the five ways to treat a risk?
Once a risk is named you have exactly five things you can do with it, and the standard framework calls the first four the methods of modifying risk. You can avoid it, by not taking the exposure at all; the circle of competence is avoidance wearing an epistemological hat. You can prevent and mitigate it, keeping the exposure while cutting the probability or the severity; diversification, a position cap and the margin of safety all live here. You can transfer it, paying an insurer to carry the loss, which requires an insurable interest and a pool of similar risks the insurer can diversify across. You can shift it, changing the payoff's shape with a derivative, which requires neither an insurable interest nor a pool but does bring basis risk, margin calls and a counterparty. Or you can retain it deliberately, which means holding capital against it and saying so. Retention is also the default for every risk you failed to name, and that is the argument for the taxonomy in one line.
Worked example: a ₹40 crore concentrated position
The two middle methods look interchangeable and are priced by different machinery, so it is worth putting one exposure through both and reading the bill. Take a ₹40 crore concentrated position in a single listed name, held for one more year, with a 28% annualised volatility, a 1.2% dividend yield and a 6.5% one-year risk-free rate, and suppose you want the loss floored at a 20% fall. Shifting the risk means buying a one-year put struck 20% below spot. On those inputs it is worth 2.048240 per 100 of exposure, so ₹81,92,961 on the position, or 2.05% of it, paid once and gone.
Transferring the same tail to an indemnity writer prices differently, because an insurer starts from its own expected loss and adds a loading. Under a real-world drift of 9.8% on the price, the probability of finishing below the floor is 15.6982% and the present value of the expected payout, the pure premium, is 1.530100 per 100. At a 1.45 loading for expenses, capital and profit the quote is 2.218644 per 100, or ₹88,74,578, which is ₹6,81,616 more than the traded put. Read the two markups side by side and the lesson is sharper than the ranking. The traded put also costs more than its own pure premium, by 33.8632%, because whoever sells you the tail wants paying for bearing it. Neither route is charity. One markup is a risk premium set by competition among writers who can hedge dynamically, and the other is a risk premium plus an insurer's cost of holding capital it cannot hedge at all.
Retention and the zero-premium collar complete the picture, and both are more expensive than they look. Retaining means holding unencumbered capital against the loss instead of buying protection. At the 99% one-year level the price falls to 55.290040 against a spot of 100, so you would set aside ₹17,88,39,841 to absorb it without ever being a forced seller, and carrying that capital at a 3.5-point spread over the cash it earns costs ₹62,59,394 a year. The collar looks free and is the dearest of the four. Selling the call that exactly funds the put puts the cap at 145.037561, meaning you keep nothing above a 45.0376% rise, and the present value of the upside you have handed away is 2.865260 per 100, or ₹1,14,61,041. That is 1.398889 times the put's certain premium, paid in a currency nobody invoices you in. A structure with no cash cost is not a structure with no cost, and the sentence to carry out of this section is that every one of the five methods is paid for, in premium, in given-up return, or in capital that has to sit still.