---
title: "Crisis Management"
book: "The Desk and the Firm"
subject: quant
language: en
chapter: 27
exercises: 8
source: https://one-course.com/books/quant/16/en/chapter/27-crisis-management
---

# Chapter 27 — Crisis Management

On 31 October 2011 liquidation proceedings began for MF Global, a large futures commission merchant, and its customers learned that money which should have been held apart for them was missing. The regulator later found that during the last week of October the firm had used customer segregated funds to support its own and its affiliates’ operations; a court ordered $1.212 billion of restitution in 2013, and customers were eventually repaid in full. In the first days, firms that cleared through it had to find out which of their positions still existed, where their margin was and who could move it. A crisis gives a firm hours, and what it can do in those hours is decided long before.

## 27.1 The first hour of a technology incident

**Definition 27.1 (Crisis management plan, crisis committee).**

A *crisis management plan* sets out, for defined kinds of crisis (a trading system out of control, a liquidity squeeze, a counterparty failure, a market closure), who decides, what they may do, whom they tell and in what order, with the runbooks and contact lists to do it. The *crisis committee* is the small group the plan names to take those decisions, with the authority to stop trading, cut positions, move collateral and draw funding without the usual approvals.

The first hour of a technology incident is Book 13’s and Book 15’s: the kill switch and the mass cancel stop the damage, the incident commander runs the response from the runbook, failover moves trading to the disaster-recovery site (Book 14) if the primary is lost. What the [crisis committee](#def-fm-crisis-management-plan) adds is the business decision around the technical one: whether to resume, which positions to flatten, what to tell counterparties, clients and the regulator, and how much cash the incident has already cost. Book 14’s business continuity plan and operational resilience rules set the recovery targets; the crisis plan decides who acts when those targets are about to be missed.

## 27.2 Liquidity squeezes: margin calls you cannot meet

A liquidity crisis is chapter 14’s stress test happening in hours instead of days: requirements rise at every broker and clearing house at once, funding sources that looked reliable are withdrawn, and assets that could be sold on a normal day take longer or cost more. The [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) is measured in hours, and every action has a delay.

**Definition 27.2 (Liquidity drill).**

A *liquidity drill* rehearses a liquidity crisis on the firm’s real accounts, terms and funding sources: a scripted scenario is run hour by hour, the [crisis committee](#def-fm-crisis-management-plan) chooses actions under their real delays and costs, and the drill records the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) and the decisions, to be compared with the plan.

## 27.3 When a counterparty fails

**Definition 27.3 (Trapped assets).**

*Trapped assets* are a firm’s cash or securities held by a counterparty that has failed or been placed in insolvency proceedings, which the firm cannot use until an administrator or trustee returns them, if ever: collateral posted, excess margin, securities rehypothecated (Book 1), or balances not segregated as the rules required.

When a prime broker or clearing broker fails, three things happen at once. The collateral it held is trapped, wholly or in part, for months or years (Book 1 recounted the administration of Lehman Brothers’ European broker-dealer). The positions it held must be moved or re-established elsewhere, and the new broker requires margin for them immediately. And the firm’s own counterparties, seeing the news, may raise their terms. Customer asset segregation (Book 3) is the protection; MF Global showed that its value depends on the broker obeying it.

**As of September 2026 — The public record of the chapter’s examples.**

**MF Global**: liquidation proceedings began on 31 October 2011; the CFTC’s complaint charged that in the last week of October 2011 the firm used customer segregated funds to support its own and its affiliates’ operations; a consent order of 8 November 2013 required MF Global Inc. to pay $1.212 billion of restitution to customers and a $100 million penalty, and the trustee later made final distributions satisfying full restitution. **Gilt market, 2022**: on 28 September 2022 the Bank of England announced temporary purchases of long-dated UK government bonds to restore orderly market conditions, after its Financial Policy Committee noted the risks to financial stability from dysfunction in that market; the auctions ran until 14 October.

## 27.4 Tutorial: seventy-two hours

**Goal.** Run a 72-hour [liquidity drill](#def-fm-crisis-management-drill) on the firm of chapter 14’s kind: a volatility shock, a prime broker that fails, collateral trapped; find the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) and the cheapest set of actions that extends it past 72 hours. **End state:** the hourly cash chart with and without the actions ([Figure 27.1](#fig-fm-crisis-management-cash)) and the plan’s steps ([Figure 27.2](#fig-fm-crisis-management-steps)).

1. **The accounts.** Prime broker A holds $44 million against a requirement of $40 million; prime broker B $33 million against $30 million; the clearing broker $22 million against $20 million. [Unencumbered cash](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-cash) is $30 million.
2. **The shock.** [House margin](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-house) rises by 60% over the first day ( `firm.treasury.house_path` , no lock-up); the clearing house raises initial margin by 30% at the first daily cycle and 50% at the second.
3. **The failure.** Prime broker B fails at hour 6: 60% of the $33 million it holds is trapped, the rest comes back at hour 48, and its positions move to prime broker A.
4. **The actions.** Sell money-market funds ($10 million after 4 hours, cost $0.02 million); repo the bond portfolio ($10 million after 8 hours, $0.1 million); draw the committed line ($20 million after 2 hours, $0.3 million, withdrawn at hour 12); cut positions by 10%, 20% or 30% (after 3 hours, at impact costs of $1, 3 and 6 million).
5. **The plan.** `firm.crisisdrill.greedy` adds the action with the lowest cost per hour gained until the horizon reaches 72 hours.

```python
def run(accounts, cash0, sc, actions=(), hours=72):
    held = {a.name: a.held for a in accounts}
    base = {a.name: a.base_req for a in accounts}
    alive = {a.name: True for a in accounts}
    cash, calls = cash0, 0.0
    path, callpath = [], []
    for h in range(hours):
        cut = sum(a.cut for a in actions if a.delay <= h and (a.deadline is None or a.delay <= a.deadline))
        for a in actions:
            if a.delay == h and (a.deadline is None or a.delay <= a.deadline):
                cash += a.cash
        if sc.failed and h == sc.fail_hour:
            alive[sc.failed] = False
            if sc.moved_to:
                base[sc.moved_to] += base[sc.failed]
        if sc.failed and h == sc.return_hour:
            cash += held[sc.failed] * (1 - sc.trapped)
        call_h = 0.0
        for n in held:
            if not alive[n]:
                continue
            mult = sc.clearing_mult[h] if n == sc.clearing else sc.house_mult[h]
            req = base[n] * mult * max(0.0, 1 - cut)
            short = req - held[n]
            if short > 0:
                held[n] += short
                cash -= short
                call_h += short
            elif short < -1e-9 and n == sc.clearing:
                held[n] += short                   # the clearing account returns excess daily variation
                cash -= short
        calls += call_h
        path.append(cash)
        callpath.append(call_h)
    path = np.array(path)
    neg = np.nonzero(path < 0)[0]
    return {"cash": path, "calls": np.array(callpath), "horizon": int(neg[0]) if len(neg) else hours,
            "total_calls": calls, "cost": sum(a.cost for a in actions)}

```

***Listing 27.1.** The hourly liquidity engine: calls paid from cash as requirements rise, the failed broker’s positions moved and its collateral trapped, and actions completing after their delays. code/firm/crisisdrill/firm_crisisdrill.py*

![Seventy-two hours of unencumbered cash under the drill: with no action the firm runs out of cash in hour 6, when the failed broker’s positions must be margined again at the surviving one; with the cheapest plan it keeps at least $11.8 million. Without the failure the shock alone exhausts the cash in hour 23. Data: fm_crisis.](https://one-course.com/images/onecourse/chapters/quant-16/fm-crisis-management/fig-d140a2451acf.svg)

***Figure 27.1.** Seventy-two hours of [unencumbered cash](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-cash) under the drill: with no action the firm runs out of cash in hour 6, when the failed broker’s positions must be margined again at the surviving one; with the cheapest plan it keeps at least $11.8 million. Without the failure the shock alone exhausts the cash in hour 23. Data: `fm_crisis`.*

Without action the firm is out of cash in hour 6. The failure does the damage: $19.8 million of collateral is trapped at the failed broker, and its positions need margin at the surviving one at once; the shock alone would have exhausted the cash in hour 23. Over the 72 hours the calls total $78.8 million. The greedy plan sells the money-market funds first (the horizon moves to hour 9 for $20 000), then repos the bonds (hour 15), then draws the line before it is withdrawn (hour 24), and only then cuts 10% of positions, which carries the firm past 72 hours: $1.42 million in all, most of it the cut’s market impact. The three cash sources together reach only hour 24; without the cut, the second daily clearing cycle would still exhaust the cash.

![The survival horizon as the greedy plan adds actions, cheapest per hour gained first; cumulative costs $0, 0.02, 0.12, 0.42 and 1.42 million. Data: fm_crisis.steps.](https://one-course.com/images/onecourse/chapters/quant-16/fm-crisis-management/fig-3704429a8466.svg)

***Figure 27.2.** The [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) as the greedy plan adds actions, cheapest per hour gained first; cumulative costs $0, 0.02, 0.12, 0.42 and 1.42 million. Data: `fm_crisis.steps`.*

**Proposition 27.4 (More resources never shorten the horizon).**

In the drill’s model, adding an action that brings cash or cuts requirements, without changing any other action, never makes the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) shorter.

**Proof.** Cash at each hour is the starting cash plus completed actions’ cash, plus returns of excess, less calls. A cash action adds a non-negative amount from its completion onwards and changes no call. A cut lowers every later requirement, so every later call is no larger (the prime brokers retain excess, and the clearing account returns it, which only adds cash). Cash is therefore at least as high at every hour, and the first negative hour is no earlier. ∎

**Remark 27.5 (Greedy is not optimal).**

Choosing by cost per hour gained is a heuristic. Hours are not additive: an action with a long delay gains nothing alone and a great deal after others have bought the time it needs, and a line that is withdrawn at hour 12 must be drawn early or never. For a plan with a handful of actions, an exhaustive search is cheap; the drill’s value is in the committee practising the choice, not in the algorithm.

## 27.5 Drills and playbooks

**Method 27.6 (Running a liquidity drill).**

1. Write the scenario from the firm’s own accounts and terms: which broker fails, what is trapped, what is withdrawn, how margins rise, hour by hour.
2. Give the [crisis committee](#def-fm-crisis-management-plan) the real contact lists, the real authorities and the real delays of each action, and run it without warning.
3. Record every decision with its time; compute the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) achieved against the plan’s.
4. Fix what the drill exposed: a line that needed two signatures nobody could reach, a transfer that missed a cut-off, a runbook that assumed a system that was down.
5. Repeat at least yearly, and after any change of broker, funding line or strategy.

The Financial Stability Board’s 2024 recommendations on liquidity preparedness for margin and collateral calls, summarised in chapter 14, ask non-bank firms for exactly this: stress tests of margin calls under extreme but plausible scenarios and plans to meet them. A drill is where the plan meets the clock.

## 27.6 Build: the crisis drill

**Purpose.** An hourly [liquidity drill](#def-fm-crisis-management-drill): scenarios with shocks, a failing counterparty, [trapped assets](#def-fm-crisis-management-trapped) and withdrawn lines; the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon); and an action plan chosen by cost per hour gained.

**Interface.** `firm.crisisdrill`: `Account`, `Scenario`, `Action`, `run`, `greedy`; the house-margin path from `firm.treasury`.

**Rules.** Calls are paid the hour they arise; a failed account stops and returns its untrapped collateral at the stated hour; an action is available only up to its deadline.

**Acceptance tests.** `code/firm/crisisdrill/tests/`: calls and horizon on hand numbers; failure, trapping and transfer; the greedy choice and a withdrawn line.

**Stretch.** Exhaustive search over action sets; uncertain delays; the clearing house’s own margin model from `firm.initmargin`; several failures.

Sources and further reading

- CFTC press releases 6776-13 (2013), 6904-14 (2014) and 7508-17 (2017) on MF Global.
- Bank of England, “Bank of England announces gilt market operation”, 28 September 2022.

## 27.7 Exercises

**Exercise 27.1 ★.**

How much collateral is trapped at the failed broker, and how much comes back at hour 48?

**Solution of Exercise 27.1.**

$0.6\times33=\$19.8$ million trapped; $0.4\times33=\$13.2$ million comes back at hour 48.

**Exercise 27.2 ★.**

Define [trapped assets](#def-fm-crisis-management-trapped) and give three ways a firm’s assets can be trapped at a failed broker.

**Solution of Exercise 27.2.**

See [Definition 27.3](#def-fm-crisis-management-trapped): collateral posted as margin, excess margin not yet returned, rehypothecated securities, and balances that should have been segregated but were not.

**Exercise 27.3 ★.**

In which hour does the firm run out of cash without the failure, and why so much later?

**Solution of Exercise 27.3.**

Hour 23: without the failure, only the shock’s gradual margin increases drain cash, and nothing needs to be re-margined at once.

**Exercise 27.4 ★★.**

Why does the greedy plan draw the committed line only third, and what would happen if the line needed two hours of approvals from a person who is unreachable until hour 12?

**Solution of Exercise 27.4.**

The money-market funds and the repo buy hours more cheaply per hour, so the greedy rule takes them first; the line is still drawn well before its withdrawal. If it could not be drawn at all, the plan would need a second 10% cut and cost $3.12 million instead of $1.42 million: an approval path that fails in a crisis is worth $1.7 million here.

**Exercise 27.5 ★★.**

Prove [Proposition 27.4](#prop-fm-crisis-management-monotone). Would it hold if cutting positions triggered losses paid in cash?

**Solution of Exercise 27.5.**

See the proof of [Proposition 27.4](#prop-fm-crisis-management-monotone). If cutting realised losses paid in cash, a cut could lower cash in the hours before it reduces calls, and the proposition would fail.

**Exercise 27.6 ★★.**

What did MF Global’s customers need to know in the first days, and which records would have told them?

**Solution of Exercise 27.6.**

Which of their positions still existed and where, how much margin they had posted and in which accounts, and whether it was segregated: the broker’s daily statements, the firm’s own position and collateral records, and the clearing house’s records of the broker’s customer accounts.

**Exercise 27.7 ★★★.**

*Coding.* Run the drill with all three cash actions and no cut. What is the horizon, and what is the lowest cash?

**Solution of Exercise 27.7.**

Hour 24, with cash at $-\$2.8$ million at its lowest: the second daily clearing cycle still exhausts it.

**Exercise 27.8 ★★★.**

*Find the flaw.* “We have $30 million of cash and $40 million of lines: we can survive any margin call.”

**Solution of Exercise 27.8.**

Lines can be withdrawn in exactly the crisis that needs them, cash can be trapped at a failing broker, and margin can rise at every counterparty at once; the drill’s calls reach $78.8 million, and the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) depends on timing, not totals.

## 27.8 Problem: Seventy-Two Hours

**Problem 27.1.**

Weekend problem — seventy-two hours

A [chief risk officer](https://one-course.com/books/quant/16/en/chapter/12-the-risk-management-function#def-fm-the-risk-management-function-cro) must run the firm’s first [liquidity drill](#def-fm-crisis-management-drill) and report what it found to the board.

**Part I — The plan.**

1. Define a [crisis management plan](#def-fm-crisis-management-plan) and a [crisis committee](#def-fm-crisis-management-plan) .
2. What does the committee add to a technology incident’s response?
3. Define a [liquidity drill](#def-fm-crisis-management-drill) .
4. Define [trapped assets](#def-fm-crisis-management-trapped) .

**Part II — The record.**

5. Summarise the public record on MF Global.
6. What did the Bank of England do in September 2022, and why?
7. What protects customer assets at a broker, and what did MF Global show about it?
8. What do the 2024 international recommendations ask of non-bank firms?

**Part III — The drill.**

9. Describe the scenario hour by hour.
10. Give the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) with no action, and without the failure.
11. Give the greedy plan’s steps, horizons and costs.
12. State and prove [Proposition 27.4](#prop-fm-crisis-management-monotone) .
13. Why is the greedy plan not necessarily the cheapest?

**Part IV — The report.**

14. Which weaknesses would a real drill likely expose that the model does not?
15. What would you change in the firm’s broker arrangements?
16. How large should the [liquidity buffer](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-cash) be?
17. How often should the drill run, and who should not know in advance?
18. What would you report to the board?
19. State the *named result* : the [survival horizon](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-horizon) under the drill, and the cost of the cheapest action set that extends it to 72 hours.
20. In two sentences, write the recommendation.

**Solution of Problem 27.1.**

1. See [Definition 27.1](#def-fm-crisis-management-plan) .
2. The business decisions around the technical fix: resume or not, what to flatten, whom to tell, what it has cost.
3. See [Definition 27.2](#def-fm-crisis-management-drill) .
4. See [Definition 27.3](#def-fm-crisis-management-trapped) .
5. See [Box 27.1](#dat-fm-crisis-management-record) .
6. It bought long-dated gilts temporarily, from 28 September to 14 October, to restore orderly market conditions, after its Financial Policy Committee noted the risks from the market’s dysfunction.
7. Customer asset segregation; it protects only if the broker obeys it and its regulator can see that it does.
8. Liquidity risk management for margin and collateral calls, stress tests of them, and [collateral management](https://one-course.com/books/quant/16/en/chapter/13-operations#def-fm-operations-collateral) with enough cash and liquid assets.
9. A 60% rise in [house margin](https://one-course.com/books/quant/16/en/chapter/14-treasury-and-funding#def-fm-treasury-and-funding-house) over the first day, clearing margin up 30% and then 50% at the daily cycles, broker B failing at hour 6 with 60% of its collateral trapped, the rest back at hour 48.
10. Hour 6; hour 23 without the failure.
11. Money-market funds (hour 9, $0.02 million), repo (hour 15, $0.12 million cumulative), line (hour 24, $0.42 million), a 10% cut (past 72 hours, $1.42 million).
12. See [Proposition 27.4](#prop-fm-crisis-management-monotone) .
13. Hours are not additive and deadlines constrain order; an exhaustive search over small action sets is exact.
14. Approvals that cannot be obtained, cut-off times missed, systems down, counterparties that do not answer.
15. More than one prime broker with lock-ups, segregation and rehypothecation limits, and excess collateral moved out daily.
16. At least the drill’s largest shortfall plus the cost of the actions that are slow or uncertain.
17. Yearly and after major changes; the committee should not know the scenario in advance.
18. The horizon achieved, the decisions and their times, what broke, and the fixes with owners and dates.
19. Six hours with no action; $1.42 million for the cheapest set that reaches 72 hours.
20. Hold money-market funds and a committed line that can be drawn within hours, spread collateral across brokers and move excess out daily; run the drill yearly and fix what it exposes before the next.

## 27.9 Interview questions

**Interview question 27.1 ★ developer.**

A strategy is sending orders it should not. What do you do in the first five minutes?

**Solution of Interview question 27.1.**

Hit the kill switch or mass cancel, stop the strategy, confirm the orders are gone, check the position and the P&L, and call the incident commander.

*What the interviewer is looking for: stop first, then assess.*

**Interview question 27.2 ★ risk.**

Your prime broker has just announced it is in administration. What do you check first?

**Solution of Interview question 27.2.**

What it holds of ours and whether it is segregated or rehypothecated, our positions and their margin elsewhere, where to move positions, and our cash for the calls that follow.

*What the interviewer is looking for: [trapped assets](#def-fm-crisis-management-trapped) and replacement margin.*

**Interview question 27.3 ★★ risk, trader.**

Margin calls exceed your cash by $10 million tomorrow morning. List your options in order.

**Solution of Interview question 27.3.**

Liquid assets that can be sold the same day, committed lines, collateral transfers from accounts with excess, repo, and cutting positions, in order of cost per hour of survival gained, checking each delay against the call’s deadline.

*What the interviewer is looking for: delays and costs, not only amounts.*

**Interview question 27.4 ★★ developer, risk.**

How would you design a [liquidity drill](#def-fm-crisis-management-drill) that tests people as well as models?

**Solution of Interview question 27.4.**

Run it without warning on real contact lists and authorities, with injected failures (a signatory absent, a system down), and record decisions and times against the plan.

*What the interviewer is looking for: surprise and realism.*

**Interview question 27.5 ★★ trader.**

Why can cutting positions in a crisis cost more than it saves?

**Solution of Interview question 27.5.**

Selling into a stressed market has a large impact, and the losses realised can exceed the margin released; others selling the same positions make it worse.

*What the interviewer is looking for: impact and crowding.*

**Interview question 27.6 ★★★ researcher, risk.**

Formulate the choice of crisis actions as an optimisation problem, and say why a greedy rule can fail.

**Solution of Interview question 27.6.**

Choose a subset of actions with delays, costs and deadlines to minimise cost subject to cash being non-negative every hour; a greedy rule ignores interactions (an action that is useless alone but decisive after others), so a small exhaustive or integer search is better.

*What the interviewer is looking for: constraints on every hour and interactions.*