What a Position Ceiling Rests On, and Who Can Move It
A position boundary matters most after the holding succeeds. The record should preserve the assumptions behind it, later revisions, capacity created, and the consequence of a breach.
By the team at Banyantree Investment Group
A practice note developed from the Credit Suisse record reconstructed in Eleven Months of Risk Reports on One Profitable Client
Credit Suisse’s Archegos account remained outside its US$20m potential-exposure limit for more than 170 business days. It failed the bank’s crash scenario and the replacement introduced after that scenario was reconsidered. The limit was later raised as the client’s rating fell; the bank calculated that roughly US$1bn of additional margin was required but declined to ask for it; and neither the proposed migration to dynamic margining nor the US$250m fallback deposit was in place before default.¹
Those were counterparty and collateral measures, so their arithmetic cannot be reused as a position-size formula. What transfers is the governance problem: what the boundary rested on, what changed, how much capacity the change created, and who approved it. Neither the board review nor the regulator's final notice shows a single document consolidating those fields. This note develops one for an investment portfolio: the position-ceiling record.
Before calculating a ceiling, decide what kind of number it will be. In Credit Suisse's framework, a limit required action or formal exception approval; a guideline informed. The measure closest to the cost of a stressed exit was a guideline. Every portfolio should decide which of its own numbers merely informs and which obliges action, and record the distinction.
Set the permitted loss first. For an ordinary, approximately linear long position, the starting calculation is simple:
Loss-based ceiling = permitted portfolio loss ÷ stressed percentage loss over the assumed exit
Permitted portfolio loss is the share of current portfolio value the portfolio is prepared to lose on one economic idea under the nominated stress and exit assumptions. It is one idea's share of the loss the whole portfolio can bear at once. The unit is the consolidated idea, including every instrument and account through which the portfolio expresses it. Say that amount is 1.5%, and the position is estimated to lose 40% before the exit is complete. The division gives 3.75%.
The formula describes a ceiling. Valuation, expected return, conviction, and opportunity cost decide how much of the available room to use. The calculation sets a stress budget. It asks whether the portfolio can carry the nominated loss if the state occurs; probability is a separate judgement.
Do the arithmetic at working precision, then approve a practical ceiling at or below the result. A 3.75% maximum might become 3.5% if ceilings are set in half-point increments. The calculation can be exact while the 40% stress remains a judgement; the record should preserve both the working and the convention.
A percentage decline works where loss is broadly proportional to position size. Options, short positions, leveraged instruments, and other nonlinear exposures need a full scenario loss at the proposed size. Where one scenario cannot describe the exposure, retain a small set of materially different stresses and respect the most binding of them.
A position sized to survive an adverse case may be smaller than one sized to maximise the gain if the thesis is right. The method states which error the portfolio would rather carry: owning too little of a winner, or too much of a mistake.
Run four constraints through the same record. The loss calculation is only the first pass. Where the constraints affect one another, solve them together. A larger position may lengthen the exit, and the longer exit may increase the stressed loss. The final ceiling is the lowest maximum weight produced by four:
Final ceiling = lowest of loss, liquidity, concentration, and mandate constraints
The loss constraint is the calculation above.
The liquidity constraint tests the proposed size against the exit horizon assumed in the stress. Daily turnover begins that analysis. The working inputs include the share of volume the portfolio can realistically take, market impact, execution slippage, and the likelihood that other holders are selling at the same time. The stress prices the move and execution cost during the assumed exit. The liquidity test asks whether an exit of that size is available at all. Say the stress assumes the exit completes in 20 trading days and the portfolio will take no more than a quarter of daily volume. A share trading US$30m a day indicates gross sale capacity of US$150m in that window before market impact and competing sellers; against a US$5bn portfolio, 3%.² That is the beginning of the analysis, before impact, and before anyone else is selling.
Liquidity belongs to the owner as well as the security. A listed share may trade heavily while its holder has little freedom to wait. Redemptions, collateral calls, cash commitments, a mandate breach, or an impatient committee can shorten the portfolio's timetable, reducing the permitted exit horizon or increasing the stressed loss. No volume figure will settle those conditions.
The concentration constraint catches positions that look separate until stress turns them into one exposure. Shared revenue, funding, regulation, ownership, factor sensitivity, or liquidity can make several names draw on the same loss budget.
Suppose the portfolio allows a 2.4% loss to one factor and applies a 35% stress to that group. The exact maximum group exposure is about 6.86%. If ceilings are recorded to one decimal place without exceeding the calculation, the approved group ceiling is 6.8%. With 4.6% already held, 2.2% remains for this name.² A name-by-name calculation would miss that boundary. A name answers to both budgets, its own and its group's, and the lower governs.
The group definition belongs in the record too, drawn from the economic exposure rather than around the amount of capacity a new position happens to require.
The mandate constraint begins with direct rules: maximum position weights, leverage limits, liquidity floors, cash requirements, and the permitted investment universe. Terms such as withdrawal rights, collateral provisions, and cash-flow commitments may not impose a percentage cap of their own. They change the loss or liquidity assumptions behind another constraint, and the record should show which terms cap exposure directly and which alter an input.
Run the example through all four. Loss permits 3.5%. The exit arithmetic permits at most 3%. Remaining factor capacity permits 2.2%. The mandate permits 5%. The ceiling is 2.2%, set by concentration. The position chosen may be smaller; the record shows which constraint bound, and why.
Before approval, reconcile the proposed position and group loss budgets with the portfolio's total loss tolerance under relevant joint stresses. Several positions can comply individually and still consume more loss capacity than the portfolio has when correlations rise. The record should show which combinations were tested and the aggregate loss they produced.
A failed portfolio check feeds back into the four constraints. If the combination exceeds the portfolio's tolerance, revise the permitted-loss or concentration allocation and run them again.
Write the assumptions beside the number. The permitted loss should move at the portfolio or mandate level. It may change after a drawdown, a change in liabilities, a revised mandate, or a material alteration to the portfolio's capital. It should not change because one successful position wants more room.
The stressed loss is remade for each position. That is where judgement enters most heavily, and where apparent precision can hide a generous assumption. A 40% stress may still assume the business stays solvent, the market stays open, the exit proceeds in order, and the rest of the portfolio stands still. Any of those may be reasonable. The record needs to show they were made.
For each position, write down:
• the economic idea and every instrument included in it;
• the impairment, event, or market state being stressed;
• the expected loss through the proposed exit;
• the exit horizon and realistic participation rate;
• market impact, slippage, currency, and financing effects where relevant;
• related positions drawing on the same group loss budget;
• the funding, redemption, and mandate conditions assumed to hold;
• the evidence relied on and the date of the judgement.
Historical drawdown is useful evidence and rarely a complete stress. It records what happened to the security, but says little about how much of this position the portfolio could have sold into that period, which other holdings would have been falling with it, or whether this vehicle would have retained the freedom to wait. A small unlevered account with patient capital owns a different exit from a geared vehicle holding the same name alongside correlated positions. The same ticker can carry a different exit in different hands, because the conditions of ownership are part of what the stress must price.
There is another reason to write the assumptions down. Stress estimates tend to drift in one direction: the longer a position is held and the better it performs, the gentler its imagined storm becomes. Each good year makes the floor look higher and the original number feel timid. Ease the 40 to 30 and the same 1.5% budget permits 5% instead of 3.75%: a third more position, created without buying a share or changing the thesis.² Test the denominator on the day it is set, before the position develops that history. If the 40 can be defended only by pointing to a past drawdown, it is a historical observation rather than a fully specified judgement. The record should explain why that experience remains relevant to this position, this portfolio, and this exit.
The arithmetic also describes the portfolio it permits. A 1.5% budget against a 40% stress gives an exact maximum of 3.75%. A fully invested long-only portfolio subject to a 3.5% single-position ceiling requires at least 29 positions, and more where another constraint binds lower.² A concentrated mandate may deliberately accept more loss from its largest ideas. That decision belongs in the numerator and on the record, rather than being reached by easing the stress.
Govern the record after approval. A position that works creates better evidence and a stronger constituency. The business may improve, the balance sheet may strengthen, liquidity may deepen. An original stress can become too severe, and a ceiling that never learns is another form of bad measurement.
The record should permit revision without letting the position revise its own rule. A proposed change should show:
• the old assumption and the proposed replacement;
• the evidence that changed;
• the old ceiling and the new ceiling;
• the additional capacity the revision creates;
• the scheduled review or event that triggered the work;
• who proposed the change, who challenged it, and who approved it;
• any interest those participants had in the additional capacity.
Record the relief before deciding what to do with it. That ordering separates the judgement about the measure from the desire to own more of the position.
One counterfactual does most of the work: would the same change be proposed if the position were still small and had not yet made money? A genuine improvement in the assumptions survives that question comfortably. Evidence discovered during a breach may still be valid. Its timing raises the standard of challenge, because a breach is exactly when a convenient interpretation becomes unusually attractive.
Changing the ceiling and using the capacity it creates are separate decisions. The person proposing the change should not be the only person entitled to approve the resulting increase in exposure. A small firm may not be able to separate every role. The record should still let someone who was not in the room reconstruct the proposal, challenge, approval, overlap, and interests involved. A ceiling protected that way remains changeable. What the protection removes is the quiet change, the one nobody has to own.
Compliance is tested against current exposure and current portfolio value. The original derivation stays in the history, but price appreciation alone can put a position through its ceiling with nothing bought. Passive drift and an active purchase may warrant different responses, but both belong in the record.
Give every breach a clock. A temporary excess can arise through price movement, cash settlement, or ordinary portfolio activity. Treating every crossing as an emergency encourages workarounds. The record must distinguish a passing excess from an exposure being renewed outside its boundary. In the companion feature, the second kind ran for more than 170 business days, renewed one reminder at a time.
Every breach entry should state:
• who owns the response;
• what action will restore compliance;
• the date by which it will happen;
• who may approve an extension;
• what evidence an extension requires;
• what happens when the date passes.
The response may be a smaller position, more cash or collateral, a hedge that caps the measured exposure, or an approved revision to the ceiling. What matters is that the consequence for missing the date is executable without reopening the original debate: no further purchases, a required reduction, or escalation to a named authority. Without one, the deadline is another meeting. Every extension carries a new fixed date, and repeated extensions escalate on their own.
Keep the current answer on one page. For an ordinary listed position, the front sheet stays short, with the analysis and decision trail behind it, so it can circulate without this note attached:
1. Position and capital base. Portfolio or mandate, consolidated idea, instruments included, decision date, and current portfolio value.
2. Permitted loss. The share assigned to this idea and, where relevant, its wider exposure group.
3. Four constraints and portfolio check. The maximum weight each produces, whether each obliges action or informs, which one binds, the joint stresses tested, and the resulting aggregate loss.
4. Final ceiling and current exposure. Approved ceiling, weight at approval, current weight, and remaining headroom.
5. Stress assumptions. Scenario set, loss, exit horizon, participation, market impact, related exposures, funding conditions, evidence, and date.
6. Review conditions. Scheduled review and pre-agreed events capable of reopening the judgement.
7. Change history. Latest change on the sheet, with the full retained history of assumptions, ceilings, relief, challenge, approval, and interests behind it.
8. Breach terms. Owner, action, deadline, extension evidence, predetermined consequence, and escalation path.
A position ceiling does a narrower job than a stop-loss. It limits exposure under stated assumptions; it cannot govern the loss realised after those assumptions fail. Joint-stress reconciliation is only as useful as the combinations selected. It cannot prove that every relevant dependency has been captured. Prices gap, liquidity disappears, investors shorten the horizon, and correlations arrive from somewhere the analysis did not name.³
The formula will vary across private assets, credit, options, and leveraged positions. The record can travel further. It still asks what exposure was accepted, which assumptions made it tolerable, what changed later, and who had authority to change the answer.
The arithmetic fits on one page. What it asks of a firm is harder: holding a ceiling through the years a position works, while the capped winner sits beside the laggard it funded and somebody reasonably asks why the best idea in the book is the one being restrained. The record will not win that argument. It forces the working, capacity created, and decision rights into the open, then preserves what changed: the exposure, the assumptions, or both. Whether the position comes down stays a decision the firm makes, on the record.
General information only. Not personal advice. Calculations are illustrative. Past performance is not indicative of future performance. This material is intended for wholesale and professional investors.
Related Featured Essay
For the full account behind this practice note, read Eleven Months of Risk Reports on One Profitable Client, which reconstructs the limits, stress tests, margin calculations, and decisions made as Credit Suisse’s exposure to Archegos kept growing.
Notes
1. The figures and sequence in the opening are drawn from the record reconstructed in the companion feature, which carries the fuller chronology and source apparatus. Prudential Regulation Authority, Final Notice: Credit Suisse International and Credit Suisse Securities (Europe) Limited, 24 July 2023, Annex A: the US$20 million potential-exposure limit; the classification of potential exposure as a limit and the single-factor crash scenarios as guidelines; the account's continuous presence on the standing list of limit excesses from April 2020, and the record at the start of March 2021 of more than 170 business days beyond the limit; the January 2021 downgrade from BB- to B+ alongside the proposal to raise the limit to US$50 million, approved in early February 2021; the February 2021 calculation that around US$1 billion of additional margin was required, and the 9 February response that asking for it would effectively ask the client to move its business; and the 8 March 2021 decision to seek migration to dynamic margining "within the next couple of weeks", with US$250 million as the fallback, together with the record that neither was obtained before default. Credit Suisse Group Special Committee of the Board of Directors, Report on Archegos Capital Management, 29 July 2021: use of the replacement Bad Week scenario from September 2020 and continued breaches thereafter; and the late-February proposal for dynamic margining and the cancelled follow-up calls. The observation that neither public document shows a single consolidated record of the ceiling's derivation and revision history is stated as an absence in those documents rather than as a finding that no such record existed anywhere inside the bank.
2. In the concentration example, a 2.4% group-loss budget divided by a 35% group stress gives exact maximum group exposure of approximately 6.86%. Under the illustrative convention used in the text, the approved ceiling is set at 6.8%, the nearest tenth at or below the calculation. Subtracting 4.6% of existing exposure leaves 2.2% of approved capacity. The single-name loss calculation of 1.5% divided by a 40% stressed percentage loss over the assumed exit gives 3.75%; a ceiling set in half-percentage-point increments is recorded at 3.5%, the nearest increment at or below the result. Dividing 100% by 3.5% gives approximately 28.6, so at least 29 positions are required to reach full investment under that maximum. In the liquidity example, a quarter of US$30m daily volume over 20 trading days indicates gross sale capacity of US$150m before market impact and competing sellers; against a US$5bn portfolio, US$150m is 3% of portfolio value. In the drift example, easing the stress from 40% to 30% raises the same 1.5% budget's ceiling from 3.75% to 5%, an increase of one third.
3. All calculations and percentages in this note are illustrative risk-budgeting examples. They are not forecasts, recommendations, Banyantree position limits, or descriptions of any portfolio's actual construction. The division formula assumes an approximately linear relationship between position size and stressed percentage loss over the assumed exit; nonlinear, leveraged, short, and path-dependent exposures require scenario analysis at the proposed size. Portfolio-level reconciliation depends on scenario design appropriate to the mandate, its holdings, and its funding conditions. It does not establish that every relevant dependency or joint stress has been captured.
