IN ONE SENTENCE — An agent lender lends a client’s securities to banks and broker-dealers on the client’s behalf, takes collateral through a tri-party agent, keeps a share of the fee, and usually promises to make the client whole if a borrower defaults.
Where it sits. At Meridian Bank, agency securities lending and collateral management sit in Securities Services, next to custody (see D01) and asset servicing (see D04). The client is a beneficial owner that already keeps its assets with Meridian: a pension fund, an insurer, a sovereign fund, a UCITS or ETF range. The securities never leave Meridian’s custody network: lending is a book entry from the client’s account to the borrower’s, and the collateral arrives in an account that Meridian’s tri-party service controls. This is the mirror image of the markets seat studied in C11. There, Meridian’s prime brokerage and repo desks borrow securities as principal, to deliver them to hedge funds or into settlement. Here, Meridian acts as agent for the lender: it does not own the securities, it does not take the fee risk, and it does not put the loan on its balance sheet.
How it earns money. Three revenue lines, in order of size for a European custodian bank:
Orders of magnitude for the market: ISLA’s report for the first half of 2025 put the global lendable universe at EUR 37.4 trillion and the value on loan at EUR 3.0 trillion at end-June 2025. ISLA’s report of 27 February 2026 put global securities lending revenue for 2025 at nearly USD 15bn, a record. For collateral, Euroclear reported in February 2026 a daily average of EUR 2.1 trillion outstanding on its collateral platform in 2025. Most of the revenue comes from a small share of securities lent at high fees; most of the work comes from the much larger GC book, which earns little but must be recalled, substituted, marked and reported like the rest.
The team. At Meridian about forty people work in the business in Paris, Luxembourg and London:
A week on the desk in late March. Monday: a mid-cap German stock becomes hard to borrow after a profit warning; the equity trader re-prices outstanding loans from 35 bp to 180 bp and reallocates new demand across the twelve clients that hold it. Tuesday: tri-party collateral management goes live with a new exclusion list for a Dutch pension fund; four hundred securities drop out of the eligible set overnight and the engine substitutes them before 9:00. Wednesday: the stewardship calendar flags forty European general meetings in May where Aurora, the client of this module’s case, will vote; programme management confirms the recall dates with the traders. Thursday: the credit committee reviews borrower limits; one hedge-fund-heavy broker-dealer has had its rating outlook cut, and its limit under the indemnified programme is reduced by a third. Friday: monthly client reports go out, with, for the first time, a section on collateral ESG metrics and on votes recalled.
WHY IT MATTERS ON THE DESK — The agent lender sells a promise as much as a service: that the client earns income on idle assets without losing its securities, its votes or its principles. Every control in this module protects one of those three.
IN ONE SENTENCE — In agency lending, the beneficial owner is the lender of record, the agent acts under a mandate, and the borrower faces the lender through the agent under a master agreement.
INTUITION — An agent lender is a letting agency for securities: it finds tenants, collects a deposit worth more than the flat, splits the rent with the landlord, and, if it offers a guarantee, pays the landlord if the tenant disappears and the deposit falls short.
Three contracts frame the business:
| Contract | Between | Main content |
|---|---|---|
| Securities Lending Authorisation Agreement | Client and agent | Mandate, lendable assets, borrowers, collateral, reinvestment, split, indemnity |
| GMSLA (ISLA) | Lender and borrower | Title transfer, manufactured income, margin, recall, default |
| Tri-party collateral service agreement | Borrower, lender and tri-party agent | Eligibility schedule, haircuts, concentration limits, daily valuation |
The GMSLA is the master agreement published by ISLA (the 2010 version for title transfer, a 2018 version for collateral by security interest); the text is available to members, and this module relies on public summaries. Under it, title to the securities passes to the borrower, who must return equivalent securities on demand or at term. The borrower “manufactures” dividends and coupons back to the lender, so the lender keeps the economic exposure, but not the vote. The lender can recall at any time; the borrower must return within the standard settlement cycle of the market. Collateral is marked to market daily and topped up if the margin falls below the agreed level, usually 102% for loans against cash or same-currency government bonds and 105% or more for equity collateral or cross-currency collateral.
The indemnity. Most agent lenders give the client a borrower-default indemnity: if a borrower defaults and the collateral, once sold, is not enough to buy back the lent securities, the agent pays the difference. The indemnity normally covers only this shortfall. It does not cover a fall in the value of reinvested cash collateral, nor losses from the client’s own guidelines. The distinction is not a detail: in 2008 the largest losses in securities lending came from cash reinvestment, not from borrower defaults. Large sovereign funds and some pension funds lend without indemnity, through their own credit teams; smaller lenders, and most mutual funds, insist on it.
Non-cash collateral (government bonds, equities from main indices, sometimes corporate bonds) dominates in Europe. It is usually held in tri-party: borrower and lender agree an eligibility schedule once, and the tri-party agent (Euroclear, Clearstream, or a custodian bank such as Meridian) selects eligible securities from the borrower’s account, values them, applies haircuts, enforces concentration limits and substitutes them as needed, several times a day. In ISLA’s H1 2025 data, non-cash collateral held in European tri-party was 46% equities, 38% government bonds and 14% corporate bonds.
Cash collateral dominates in the United States. The lender receives cash, pays the borrower an interest rate called the rebate, and reinvests the cash. The lender’s income is the reinvestment return minus the rebate; a specials loan has a low or negative rebate, a GC loan a rebate close to the overnight rate.
FORMULA
is the gross programme revenue a year; the value on loan in segment against non-cash collateral and its annual fee; the cash collateral received; the reinvestment yield; the rebate paid to the borrower; the client’s share of revenue (80% at Meridian).
Tri-party is a general utility. The same engine that holds a borrower’s collateral for a lender also holds repo collateral (see C11) and, since the reform of margin on derivatives, initial margin on non-centrally cleared derivatives.
Margin reform. After 2008, the G20 required margin on non-centrally cleared derivatives under the BCBS-IOSCO framework, implemented in the EU through EMIR and its delegated regulation on risk mitigation. Variation margin applies to all in-scope counterparties; initial margin was phased in from 2016 to September 2022 by size of derivatives book, the last phase covering groups above EUR 8bn of average aggregate notional. Initial margin must be segregated, which in practice means a third-party custodian or a tri-party agent, and it cannot be re-used. Collateral demand rose accordingly. More recently, the Financial Stability Board published on 10 December 2024 its final policy recommendations on the liquidity preparedness of non-bank participants for margin and collateral calls, after the 2020 dash for cash and the 2022 UK pension fund episode; BCBS, CPMI and IOSCO followed with three final reports on margin practices in January 2025. In the EU, EMIR 3 (Regulation (EU) 2024/2987) added an active account requirement at EU CCPs and changed the validation of initial margin models.
Collateral optimisation. A collateral giver holds many assets and owes collateral to many takers. Optimisation allocates the “cheapest to deliver” assets first, meaning those with the lowest opportunity cost: assets that are hard to finance elsewhere (equities, lower-rated bonds) rather than high-quality government bonds that could be lent or repo’d at a good price, within the haircuts and concentration limits of each schedule. Tri-party agents run optimisation algorithms across all of a client’s exposures; the savings are real and they depend on what each schedule accepts. That is why ESG screens, which narrow eligibility, have a cost.
INTERACTIVE LAB — Collateral optimisation: delivering the cheapest eligible assets first
Compare the annual cost of covering a collateral requirement with government bonds only and with an optimised allocation that uses equities and corporate bonds up to the schedule’s concentration limits and the available inventory.
| Parameter you can move | Range | Default |
|---|---|---|
| Collateral value required (after haircuts) | 10 to 1000 EUR m | 100 EUR m |
| Haircut on government bonds | 0 to 10 % | 2 % |
| Haircut on corporate bonds | 0 to 20 % | 8 % |
| Haircut on main-index equities | 5 to 30 % | 15 % |
| Opportunity cost of government bonds | 0 to 50 bp p.a. | 15 bp p.a. |
| Opportunity cost of corporate bonds | 0 to 50 bp p.a. | 6 bp p.a. |
| Opportunity cost of equities | 0 to 50 bp p.a. | 3 bp p.a. |
| Concentration limit on equities (share of value) | 0 to 100 % | 40 % |
| Concentration limit on corporate bonds | 0 to 100 % | 30 % |
| Eligible equity inventory available (market value) | 0 to 1000 EUR m | 100 EUR m |
| Result at default settings | Value |
|---|---|
| Annual cost, government bonds only | 0.1531 EUR m |
| Annual cost, optimised allocation | 0.0796 EUR m |
| Saving from optimisation | 48.0 % |
| Market value delivered, optimised | 110.28 EUR m |
Try it. Cut the equity inventory to zero, as a strict ESG screen on a carbon-heavy inventory could: the saving falls from about 48% to about 17%. The cost of a screen is the optimisation it forbids.
On a requirement of EUR 100m, delivering government bonds only costs about EUR 153,000 a year in forgone use of those bonds; an optimised allocation that uses equities and corporate bonds up to the schedule’s limits costs about EUR 80,000, a saving of 48%. On large books the sums are material, which is why borrowers fight over the wording of eligibility schedules.
Fees. The fee of a loan depends on scarcity: the ratio of borrowing demand to lendable supply, corporate events (dividends, rights issues, index changes), and the term. GC equities from main indices in Europe typically earn a few to around 25 bp a year; government bonds a few basis points; specials much more. Data vendors publish daily fee distributions, and clients benchmark their agent against them.
Allocation. An agent lends for many clients holding the same security. Demand must be allocated fairly, usually by a documented queue (pro rata to lendable, or rotation), audited by compliance. This “fair allocation” duty is a recurrent topic of client due diligence.
Recall. A recall instructs the borrower to return the securities. The borrower returns them from its own inventory, borrows elsewhere, or buys them; if it fails, the lender can buy them in at the borrower’s cost. The return settles in the standard settlement cycle, two business days in the EU today and one from 11 October 2027. D04 shows how to compute a recall deadline from a record date.
| Risk | What happens | Main mitigant |
|---|---|---|
| Borrower default | Borrower fails to return; collateral sold to buy back | Margin, daily marking, indemnity |
| Collateral shortfall | Lent security rises while collateral falls | Haircuts, correlation limits |
| Cash reinvestment | Reinvested assets lose value or become illiquid | Guidelines, short tenors |
| Operational | Missed recall, wrong allocation, failed substitution | Controls, reconciliations |
| Legal | Unenforceable netting or title transfer | Legal opinions per jurisdiction |
The client’s guidelines, in the authorisation agreement, fix: lendable assets and limits on utilisation; approved borrowers and borrower limits; collateral schedule; reinvestment guidelines; recall policy; and reporting. Each of these is where a sustainability preference can be written in.
IN ONE SENTENCE — Sustainability enters agency lending through four doors: the vote, the borrower, the collateral and the cash, and none of them makes the loan itself a sustainable investment.
A beneficial owner with a stewardship policy and an exclusion list asks its agent four questions. Will I be able to vote my shares at the meetings that matter? To whom are my shares lent, and for what? Will I ever hold, after a default, collateral that my own policy excludes? Where is my cash collateral invested? Each question has a contractual answer, a systems answer and a cost.
GASLA Global Framework for ESG and Securities Lending. First published in May 2021 by the Risk Management Association and PASLA with ISLA’s support, and updated in March 2023 under the Global Alliance of Securities Lending Associations, the framework addresses five areas: voting rights, collateral, lending over record dates, the facilitation of short positions, and transparency in the lending chain. It gives practical guidance for lenders rather than binding rules. Its core message is that lending and responsible ownership are compatible when the lender keeps the ability to recall, chooses its collateral deliberately and knows its counterparties.
Global PSSL. The Global Principles for Sustainable Securities Lending, an industry initiative, published a draft in July 2021 and nine principles with 17 opening signatories in September 2021. They cover, among other things, alignment with sustainable finance agendas, transparency about lending activity, the reinvestment of cash collateral in securities consistent with an ESG-compliant collateral basket where possible, and the coordination of voting policy with lending, recognising its impact on revenue. Uptake is limited and voluntary; this module uses the principles as a checklist, not as a standard that clients must certify.
ISLA on ESG collateral. In June 2022, ISLA published a paper on the applicability of ESG to collateral. Its position deserves to be learnt because it states the trade-off honestly: collateral is first a risk-management tool and must stay diversified and liquid; overly narrow ESG eligibility could harm liquidity and financial stability; and securities lending transactions and their collateral are not sustainable investments in their own right, even if they can reflect the lender’s preferences. ISLA asked regulators to confirm that collateral need not be screened as strictly as a fund’s long portfolio.
Recall for votes. Shares on loan are voted by whoever holds them at the record date. A lender that wants to vote recalls in time. Three policies exist: recall for every meeting (expensive in operations, rarely used), recall for flagged meetings (contested items, climate resolutions, meetings above a holding threshold), or never lend shares the fund intends to vote. Agent lenders now offer automated recall: the stewardship calendar flags meetings, the recall engine issues notices at the computed deadline, and the client report lists recalled positions and votes cast. Clearstream, for example, advertises automated recalls ahead of security events so that holders can exercise rights and engage. D04 computes the recall deadline for a single meeting; this module prices the policy across a programme.
Restricted borrowers and transparency. Some lenders exclude borrowers they consider inconsistent with their policies, or require borrowers to confirm that borrowed shares will not be borrowed for the purpose of voting. Since an agent cannot know the final use of most loans (most borrowing covers short sales by the borrower’s clients or settlement needs), restrictions bite on counterparties, not on uses. The cost is lost demand when restricted borrowers are significant in a security.
Short selling. Lending enables short selling. Short sellers improve price discovery and have exposed overvalued and sometimes greenwashing companies, but some owners object to supporting bets against companies they engage with. The EU Short Selling Regulation (Regulation (EU) No 236/2012) makes net short positions in shares notifiable to the national authority from 0.1% of issued capital and public from 0.5%, which gives lenders a view of aggregate short interest in a name, not of who borrowed their shares.
An ESG schedule narrows what the borrower may deliver. The tri-party agent applies the filters every day to every allocation. Market practice offers three tiers, in rising order of cost:
| Tier | Typical content | Data needed |
|---|---|---|
| 1. Norms and controversial exclusions | Controversial weapons, tobacco, UN Global Compact violators, thermal coal above a revenue threshold | Exclusion lists |
| 2. Paris-aligned Benchmark exclusions | Tier 1 plus fossil-fuel revenue thresholds and high-carbon power (Article 12 of Delegated Regulation (EU) 2020/1818) | Revenue breakdowns by activity |
| 3. Positive screens | ESG rating floor, green bond flag, best-in-class | ESG ratings, bond labels |
Tri-party agents have built these options. Euroclear offers eligibility by ESG index (it listed 65 indices in a first phase) and has worked on security-level criteria and bond flags; Clearstream offers flexible exclusion screening of collateral baskets and positive ESG bond metrics for eligibility, haircuts and concentration limits. Exclusion lists are often taken from the client’s own portfolio policy, so that the client never holds, even after a default, what it would not buy. Ratings-based screens are more contested: ESG ratings from different providers disagree widely (Roncalli’s notebook, chapter 2, explains why scores diverge), so a rating floor can exclude a security with one vendor and admit it with another.
Reinvestment guidelines specify permitted instruments (reverse repo, money market funds, government bills, sometimes commercial paper and floating rate notes), maximum maturities, credit quality and liquidity. A sustainable lender can add an ESG layer: reinvestment only in money market funds with an ESG label or exclusion policy, or reverse repo only against screened collateral. The return cost is usually small; the governance question is larger, because reinvestment risk sits with the client, not with the agent.
| Text | What it requires | Effect on the desk |
|---|---|---|
| SFTR, Regulation (EU) 2015/2365 | Reporting of each loan to a trade repository; fund disclosure of SFT use (Arts 13-14); re-use consent (Art 15) | Full trade reporting; fund reports show lending and collateral |
| ESMA guidelines on ETFs and other UCITS issues (ESMA/2014/937) | Revenues from lending, net of direct and indirect operational costs, returned to the UCITS; fees disclosed | Agent’s split must be disclosed and justified |
| SRD II, Directive (EU) 2017/828 | Engagement policy and voting disclosure for asset owners and managers | Recall policies, vote reporting |
| SFDR, Regulation (EU) 2019/2088, and ESMA fund-name guidelines | Consistency between fund characteristics and holdings | Exclusion-based collateral schedules |
| Short Selling Regulation (EU) No 236/2012 | Notification and disclosure of net short positions | Transparency on short interest |
| CRR3, Regulation (EU) 2024/1623 | Counterparty and operational-risk capital, leverage ratio | Cost of the indemnity (section 6) |
| CSDR and EU T+1 (11 October 2027) | Settlement discipline; shorter cycle | Recall deadlines one day earlier |
Three dated developments matter in November 2026. SFDR review: the Commission proposed a revised SFDR in November 2025 with product categories (Sustainable, Transition, ESG Basics); the Council agreed its negotiating mandate on 24 June 2026, the Parliament’s ECON committee adopted its report on 10 September 2026 and the plenary its mandate on 5 October 2026; the first trilogue was scheduled for 22 October 2026 and no agreement had been reached by mid-November 2026. Application is expected around 2029. The proposed Sustainable category would apply the Paris-aligned Benchmark exclusions, which reinforces their use as the default exclusion set in collateral schedules. Central banks: the ECB has applied a climate factor to the collateral value of non-financial corporate bonds since 15 June 2026 (see C11); private collateral takers increasingly look at central bank practice. United States: SEC Rule 10c-1a requires reporting of securities loans to FINRA for public dissemination; by an order of 3 December 2025 the SEC deferred reporting compliance to 28 September 2028 and public dissemination to 29 March 2029.
ESG options are now standard in the offer of large agent lenders and tri-party agents: exclusion screens, recall automation tied to a stewardship calendar, ESG metrics on collateral in client reports. What is not standard is the depth of the screen and who pays for it. Most programmes stop at norms-based or Paris-aligned exclusions; rating floors remain rare. Global PSSL has not become a market standard comparable to the ICMA principles in bonds, and no regulator has defined “sustainable securities lending”. The honest summary: the sustainable layer here is about governance of rights and of residual holdings, with a measurable cost, not about financing new green assets.
PITFALL — Do not call a lending programme “green” because its collateral is screened. The client only holds the collateral if a borrower defaults; the screen protects its policy consistency, not the climate.
The “deal” in agency lending is not a single loan: it is the set-up or redesign of a client programme, followed by daily operation and an annual review. The flow below follows the Aurora redesign of section 5, from the request to the first annual review.
INTERACTIVE LAB — From mandate to first annual review: redesigning an agency lending programme
Follow the roles, documents and controls involved in redesigning an indemnified lending programme with a stewardship overlay and an ESG tri-party collateral schedule.
| # | Who | Does what | Deliverable |
|---|---|---|---|
| 1 | Client investment committee and stewardship team | Approve a responsible lending policy: recall rules, borrower restrictions, collateral exclusions, reinvestment limits | Board-approved lending policy |
| 2 | Programme manager (agency lending) | Translate the policy into programme parameters and run a revenue impact study with the traders | Impact study: revenue by option |
| 3 | Tri-party collateral management and ESG data team | Map the exclusion tiers to vendor data, code test schedules, measure the share of borrowers’ inventory that fails | Eligibility simulation by borrower |
| 4 | Credit risk | Assess borrower limits, collateral sufficiency under the new schedule and the indemnity exposure; propose capital and EL | Credit memo for the indemnified programme |
| 5 | Finance (business line controller) | Compute agent revenue, costs, capital and RAROC; propose split and ESG service fee | Pricing note and RAROC |
| 6 | New business and pricing committee | Approve split, indemnity terms, ESG service fee and any exception to the hurdle | Committee minutes |
| 7 | Legal (client and borrower side) | Amend the Securities Lending Authorisation Agreement; update borrower GMSLA annexes and tri-party schedules | Signed SLAA amendment; tri-party collateral profile |
| 8 | Tri-party collateral management | Load the schedule; run substitutions of ineligible collateral within the agreed grace period | Go-live report; substitution log |
| 9 | Agency lending traders | Apply borrower restrictions in the allocation engine; reprice where demand falls | Updated borrower list and allocation queue |
| 10 | Proxy voting team and recall desk | Load Aurora’s flagged meetings; issue recall notices at computed deadlines; confirm returns before record dates | Recall calendar; recall log; vote confirmation |
| 11 | Cash reinvestment team | Reinvest cash collateral within the revised guidelines; daily liquidity and credit checks | Reinvestment report |
| 12 | Programme manager | Send monthly report: revenue, utilisation, recalls, collateral ESG metrics, exceptions | Monthly client report |
| 13 | Client oversight and Meridian internal audit | Annual review of performance against benchmarks, fair allocation, schedule breaches and policy fit | Annual review and policy update |
Policy and impact study (steps 1 and 2). The client owns the policy; the agent owns the arithmetic. A good programme manager never accepts a policy without an impact study: what share of the book will be off loan around meetings, what share of demand comes from restricted borrowers, what fee the borrowers will offer under the new schedule. The study is a spreadsheet built with the traders, who know borrower behaviour, and it is the basis of section 5.
Eligibility simulation (step 3). The tri-party team runs the proposed schedule against a recent snapshot of each major borrower’s collateral inventory: how much of what they delivered last month would fail? The ESG data team checks coverage: a screen applied to a security without data must default to “eligible” or “ineligible”, and that choice should be written down. Unscreened securities are a frequent audit finding.
Credit and pricing (steps 4 to 6). Credit risk looks at the indemnity: borrower limits, wrong-way risk (a borrower posting its own shares or shares of related entities is excluded), collateral sufficiency under stress. Finance computes the RAROC (section 5.7). Because the indemnity uses counterparty capital, the new business committee, not only the business head, approves the split. The committee also approves any ESG service fee and the wording of client reporting on sustainability, which compliance reviews for greenwashing.
Documentation and go-live (steps 7 to 9). The Securities Lending Authorisation Agreement is amended with the client’s guidelines: borrower list, recall policy and service level, collateral schedule by reference, reinvestment guidelines, indemnity scope. Borrowers are notified of the new schedule with a grace period to substitute collateral. On go-live, the tri-party engine flags every ineligible piece of collateral and substitutes it, or asks the borrower to do so. Controls: four-eyes on schedule coding, a test allocation in a parallel environment, reconciliation of the first day’s allocations.
Operation (steps 10 to 12). Recalls follow the stewardship calendar; the recall desk tracks returns against the record date and escalates fails. Cash reinvestment is checked daily against limits. The monthly report shows what the client pays for its policy: revenue forgone by recalls and restrictions, collateral ESG metrics, exceptions.
Review (step 13). Once a year, the client reviews the agent against fee benchmarks from market data vendors and checks fair allocation. Meridian’s internal audit tests schedule coding and data coverage. The policy is updated, and the cycle starts again.
WHO DOES WHAT — The programme manager is the hinge: client-facing, but accountable internally for revenue, for the accuracy of what the client is told about its ESG overlay, and for escalating any request that changes the indemnity risk.
Question. What does Aurora want, and what will Meridian have to price?
Data (case assumptions). Aurora Pension Fund (fictional) is a French occupational pension fund with EUR 30bn of assets. Meridian is its global custodian and, since 2019, its agent lender under an indemnified programme with an 80/20 split. Aurora’s lendable assets at Meridian are EUR 7,000m of developed-market equities and EUR 5,000m of euro-area government bonds. In September 2026 its board adopted a responsible ownership policy that applies the Paris-aligned Benchmark exclusions to its whole equity portfolio and asks for active voting at European general meetings with climate or governance items. The investment committee then asked Meridian to redesign the lending programme so that:
The same committee also asked a blunter question: is the lending income still worth the trouble, and is Meridian’s 20% justified?
| Programme parameter (today) | Value |
|---|---|
| Equities: lendable, utilisation, average fee | EUR 7,000m, 12%, 25 bp |
| Share of equity loans against cash collateral | 25%, margin 102% |
| Rebate on cash loans; reinvestment yield today | 2.25%; 2.45% (overnight reverse repo and money market funds) |
| Equity loans against non-cash collateral | Margin 105%; collateral 50% government bonds, 50% equities |
| Government bonds: lendable, utilisation, fee | EUR 5,000m, 30%, 8 bp; collateral government bonds at 102% |
| Split client / agent | 80% / 20% |
The rebate of 2.25% reflects the market: with overnight rates near 2.50% (course convention), a GC equity loan worth 25 bp is paid by a rebate 25 bp below the overnight rate.
WHO DOES WHAT — Who works on the request at Meridian: the programme manager (owner of the response), the equity and fixed income traders (revenue study), tri-party collateral management and the ESG data team (screen simulation), cash reinvestment (guidelines), credit risk (indemnity), Finance (RAROC), and the proxy voting team of D04 (recall calendar).
Question. What does the programme earn today, and how is it split?
Calculation. Equities on loan: = EUR 840m, of which EUR 210m against cash (25%) and EUR 630m against non-cash collateral. Non-cash equity loans earn = EUR 1.575m. Cash loans bring cash collateral of = EUR 214.2m, reinvested at 2.45% while Aurora pays the borrower a rebate of 2.25%: a spread of 20 bp, EUR 0.428m. Government bonds on loan: = EUR 1,500m at 8 bp, EUR 1.2m. Gross revenue: EUR 3.203m. Aurora receives 80%, EUR 2.563m; Meridian 20%, EUR 0.641m.
INTERACTIVE LAB — Aurora’s lending programme: revenue by segment and split
Compute the gross revenue of an agency lending programme from lendable assets, utilisation, fees and cash reinvestment spread, and split it between the beneficial owner and the agent lender.
| Parameter you can move | Range | Default |
|---|---|---|
| Lendable equities | 0 to 20000 EUR m | 7000 EUR m |
| Equity utilisation | 0 to 40 % | 12 % |
| Average equity fee (non-cash loans) | 1 to 200 bp p.a. | 25 bp p.a. |
| Share of equity loans against cash | 0 to 100 % | 25 % |
| Cash collateral margin | 100 to 110 % | 102 % |
| Reinvestment yield | 0 to 6 % | 2.45 % |
| Rebate paid to borrowers | -1 to 6 % | 2.25 % |
| Lendable government bonds | 0 to 20000 EUR m | 5000 EUR m |
| Bond utilisation | 0 to 60 % | 30 % |
| Average bond fee | 0 to 50 bp p.a. | 8 bp p.a. |
| Client share of revenue | 50 to 95 % | 80 % |
| Result at default settings | Value |
|---|---|
| Equities on loan | 840.0 EUR m |
| Cash collateral to reinvest | 214.2 EUR m |
| Revenue, equities against non-cash | 1.5750 EUR m |
| Revenue, equities against cash (reinvestment spread) | 0.4284 EUR m |
| Revenue, government bonds | 1.2000 EUR m |
| Gross programme revenue | 3.2034 EUR m |
| Client’s revenue | 2.5627 EUR m |
| Agent lender’s revenue | 0.6407 EUR m |
| Client’s revenue on total lendable | 2.14 bp |
| Gross revenue on value on loan | 13.69 bp |
| Share of revenue from government bonds | 37.5 % |
Try it. Double equity utilisation to 24%: revenue rises by 63%, not 100%, because bonds are unchanged. Then set the client share to 75%: the agent’s revenue rises by a quarter, the client’s falls by about 6%.
Result. Gross revenue of EUR 3.203m on EUR 2,340m on loan is 13.7 bp on the value lent; Aurora’s EUR 2.563m is 2.1 bp on its EUR 12bn of lendable assets.
Reading. Two points shape the rest of the case. First, the income is small relative to Aurora’s portfolio: 2.1 bp a year is roughly what Aurora pays Meridian for custody. That is why every ESG overlay has to be weighed against a thin base. Second, more than half of the revenue comes from equities, which is exactly where the stewardship and collateral questions arise; government bond lending, the other 37%, is largely untouched by the policy.
PITFALL — Do not compare revenue on value on loan (13.7 bp) with revenue on lendable assets (2.1 bp) across agents without saying which one you use. Agents marketing their performance like the first; trustees should look at the second.
Question. Aurora’s asset manager suggests “enhanced” reinvestment to raise income. What does it earn, what can it lose, and what does an ESG filter cost?
Data (case assumptions). The cash pool is EUR 214.2m (stage 5.2). Conservative guidelines (overnight reverse repo against government bonds, euro money market funds) yield 2.45%. Enhanced guidelines (commercial paper and floating rate notes of banks and corporates up to one year) yield 2.70%. In a stress like 2008, the enhanced portfolio could lose 2% of its value through credit losses and forced sales. An ESG filter on the conservative guidelines (money market funds applying exclusions, reverse repo against screened collateral) costs 2 bp of yield. Reinvestment losses are borne by Aurora alone: the borrower-default indemnity does not cover them.
FORMULA
is the extra annual revenue of the enhanced guidelines; the cash pool; and the two reinvestment yields; the number of years of the client’s extra income that one stress loss wipes out; the loss rate in stress; the client’s share of revenue.
INTERACTIVE LAB — Cash collateral reinvestment: extra spread versus loss in stress
Compare conservative and enhanced reinvestment guidelines by extra income and by the number of years of that income consumed by one stress loss, and measure the cost of an ESG filter.
| Parameter you can move | Range | Default |
|---|---|---|
| Cash collateral pool | 10 to 5000 EUR m | 214.2 EUR m |
| Rebate paid to borrowers | 0 to 6 % | 2.25 % |
| Yield, conservative guidelines | 0 to 6 % | 2.45 % |
| Yield, enhanced guidelines | 0 to 7 % | 2.7 % |
| Loss of the enhanced portfolio in stress | 0 to 10 % | 2 % |
| Yield cost of the ESG filter (conservative) | 0 to 10 bp | 2 bp |
| Client share of revenue | 50 to 95 % | 80 % |
| Result at default settings | Value |
|---|---|
| Reinvestment revenue, conservative | 0.4284 EUR m |
| Reinvestment revenue, enhanced | 0.9639 EUR m |
| Client’s extra income from enhanced guidelines | 0.4284 EUR m |
| Stress loss, borne by the client | 4.284 EUR m |
| Years of the client’s extra income lost in one stress | 10.0 years |
| Annual cost of the ESG filter | 0.0428 EUR m |
| ESG filter cost as share of conservative revenue | 10.0 % |
Try it. Set the stress loss to 5%, a severe credit and liquidity stress: one bad year erases 25 years of enhanced income. Then set the enhanced yield to 2.55%: the trade-off becomes absurd.
Result. Enhanced guidelines raise reinvestment revenue from EUR 0.428m to EUR 0.964m. Aurora’s share of the extra is EUR 0.428m a year. One stress loss of 2% costs Aurora EUR 4.284m, ten years of that extra income. The ESG filter on the conservative guidelines costs EUR 42,840 a year, 10% of reinvestment revenue.
Reading. The enhanced option is a short put on credit spreads and liquidity, sold for 25 bp. It also changes the nature of the programme: the client is no longer lending securities, it is running a leveraged short-term credit fund financed by rebates. The 2008 experience is the reference here: AIG’s lending programme had reinvested cash collateral in long-dated mortgage-backed securities and needed Federal Reserve support (the Maiden Lane II vehicle took over the assets). Meridian’s recommendation is to keep the conservative guidelines and add the ESG filter. Its cost is small, its effect on Aurora’s exposure is nil, and it aligns with the Global PSSL principle on reinvestment. The filter does not make the cash “green”; it avoids lending Aurora’s cash, through reverse repo or fund holdings, to issuers Aurora excludes.
NOTE — Why does the indemnity exclude reinvestment? Because the agent does not choose the risk: the guidelines are the client’s. An agent that also managed the reinvestment, and indemnified it, would be guaranteeing its own investment decisions, a different and much larger promise.
Question. What do Aurora’s recall policy and borrower restrictions cost the programme, and who pays?
Data (case assumptions). Equity revenue is EUR 2.003m (stage 5.2: EUR 1.575m non-cash plus EUR 0.428m cash). Companies with flagged meetings represent 40% of equities on loan. Under the recall policy the shares are off loan 10 days a year on average; because borrowing demand and fees spike around record dates of contested meetings, the fee forgone on those days is twice the average (spike factor 2). Borrowers that do not give the voting confirmation account for 15% of equity demand; Meridian’s traders expect to place 60% of that demand with other borrowers, so 40% of it is lost. The alternative policy, used by some funds, is never to lend the shares they intend to vote.
FORMULA
is the revenue lost to recalls; the equity revenue; the share of equities on loan with flagged meetings; the fee spike factor around record dates; the days off loan per year; the revenue lost to borrower restrictions; the share of demand from restricted borrowers; the share of that demand the agent can reallocate.
INTERACTIVE LAB — The cost of a stewardship overlay: recalls and restricted borrowers
Compute the revenue lost by recalling shares for flagged meetings and by excluding borrowers, compare it with never lending voted shares, and split the cost between client and agent.
| Parameter you can move | Range | Default |
|---|---|---|
| Equity revenue of the programme | 0.1 to 20 EUR m | 2.0034 EUR m |
| Gross programme revenue | 0.1 to 30 EUR m | 3.2034 EUR m |
| Share of equities on loan with flagged meetings | 0 to 100 % | 40 % |
| Days off loan per year (recall policy) | 0 to 60 days | 10 days |
| Fee spike factor around record dates | 1 to 5 x | 2 x |
| Share of demand from restricted borrowers | 0 to 50 % | 15 % |
| Share of that demand reallocated | 0 to 100 % | 60 % |
| Client share of revenue | 50 to 95 % | 80 % |
| Result at default settings | Value |
|---|---|
| Revenue lost to recalls | 0.0439 EUR m |
| Revenue lost to borrower restrictions | 0.1202 EUR m |
| Total cost of the overlay | 0.1641 EUR m |
| Overlay cost as share of gross revenue | 5.12 % |
| Borne by the client | 0.1313 EUR m |
| Borne by the agent | 0.0328 EUR m |
| Revenue lost if voted shares are never lent | 0.8014 EUR m |
| Withholding cost as a multiple of the overlay | 4.88 x |
Try it. Set reallocation to 100%: restrictions cost nothing, which is what happens when the excluded borrowers are small. Set it to 0% on a specials-heavy book and the restriction becomes the most expensive part of the policy.
Result. Recalls cost EUR 43,910 a year, borrower restrictions EUR 120,204: a total of EUR 164,114, 5.1% of gross revenue. Aurora bears EUR 131,291 and Meridian EUR 32,823 through the split. Withholding the voted shares from lending would cost EUR 801,360, almost five times as much (4.9 times).
Reading. The policy most visible to Aurora’s board, recalling for votes, is the cheapest. The restriction on borrowers costs nearly three times more and its benefit is harder to show: excluded borrowers can still borrow the same shares from other lenders. Meridian suggests two refinements. Replace the outright exclusion with a contractual representation in the borrowers’ GMSLA annex that shares are not borrowed for the purpose of voting, which most large broker-dealers already give. And measure recall effectiveness (shares returned before record date over shares recalled) in the monthly report; a policy that is cheap but fails in 10% of cases protects nothing. The arithmetic matches C11 (where Lyra’s recall policy cost about 1% of its income) and D04 (a single meeting): recalling is cheap, withholding is not.
PITFALL — The spike factor matters more than the days. For a contested meeting with heavy borrowing demand, fees can multiply around the record date, and that is also when an empty vote is most valuable to a borrower. A policy priced on average fees understates its own cost and its own value.
Question. Which ESG tier should Aurora’s collateral schedule apply, and what does each tier cost?
Data (case assumptions). Screens apply to the equity collateral received against equity loans; government bond collateral is not screened (Aurora’s policy applies to corporate issuers). Non-cash equity loans are EUR 630m at a 105% margin, half of the collateral being equities. Meridian’s tri-party team ran each tier against the last month’s inventory of Aurora’s six main borrowers:
| Tier | Additional share of equity collateral excluded | Funding differential of the replacement |
|---|---|---|
| 1. Norms and controversial exclusions | 8% | 20 bp |
| 2. Paris-aligned Benchmark exclusions | 10% | 30 bp |
| 3. ESG rating floor (bottom quintile out) | 15% | 45 bp |
The funding differential rises with each tier because borrowers replace excluded equities first with what is cheapest to them (other equities), then with assets they would rather use elsewhere (government bonds). Borrowers pass the full cost on to the lender through a lower fee, so Aurora and Meridian bear it in proportion to the split.
FORMULA
is the annual cost of applying tiers 1 to ; the value of non-cash equity loans; the margin; the share of equities in the collateral; the additional share of equity collateral excluded by tier ; the funding differential of the replacement for tier .
INTERACTIVE LAB — Tiered ESG screens in a tri-party schedule: cost by tier
Compute the annual cost of each cumulative ESG eligibility tier for the collateral received against equity loans, its share of programme revenue, who bears it, and how much collateral moves into government bonds.
| Parameter you can move | Range | Default |
|---|---|---|
| Equity loans against non-cash collateral | 50 to 5000 EUR m | 630 EUR m |
| Collateral margin | 100 to 115 % | 105 % |
| Share of equities in the collateral | 0 to 100 % | 50 % |
| Tier 1: additional share excluded | 0 to 50 % | 8 % |
| Tier 1: funding differential | 0 to 100 bp | 20 bp |
| Tier 2: additional share excluded | 0 to 50 % | 10 % |
| Tier 2: funding differential | 0 to 100 bp | 30 bp |
| Tier 3: additional share excluded | 0 to 50 % | 15 % |
| Tier 3: funding differential | 0 to 100 bp | 45 bp |
| Tiers applied (cumulative) | 0 / 1 / 2 / 3 | 2 |
| Gross programme revenue | 0.1 to 30 EUR m | 3.2034 EUR m |
| Client share of revenue | 50 to 95 % | 80 % |
| Result at default settings | Value |
|---|---|
| Equity collateral subject to screens | 330.75 EUR m |
| Annual cost of the chosen tiers | 0.1521 EUR m |
| Cost in bp of equity loans | 2.42 bp |
| Cost as share of gross revenue | 4.75 % |
| Borne by the client | 0.1217 EUR m |
| Borne by the agent | 0.0304 EUR m |
| Equity collateral replaced (mostly by government bonds) | 59.535 EUR m |
| Cost if all three tiers are applied | 0.3754 EUR m |
| All three tiers as share of gross revenue | 11.72 % |
Try it. Raise tier 3’s funding differential to 80 bp, as happens when borrowers are short of government bonds at quarter-end: the rating floor alone then costs more than the rest of the overlay together.
Result. Equity collateral subject to screens is EUR 330.75m. Tier 1 costs EUR 52,920 a year; tiers 1 and 2 together EUR 152,145, 2.4 bp of equity loans and 4.75% of gross revenue (Aurora EUR 121,716, Meridian EUR 30,429). Adding the rating floor takes the cost to EUR 375,406, 11.7% of gross revenue. The recommended schedule (tiers 1 and 2) moves EUR 59.5m of collateral from equities to government bonds.
Reading. The cost is convex: the last tier excludes the most and costs the most per euro excluded. Meridian recommends tiers 1 and 2 for three reasons. They mirror Aurora’s own portfolio exclusions, so the policy argument (never hold what you would not buy) is complete. They use data with good coverage (activity revenue breakdowns), unlike ratings that disagree across vendors. And ISLA’s 2022 position, that collateral should stay diversified and liquid, argues against the narrowest screens. Two further design points come from the tri-party desk:
Finally, the shift towards government bonds has a side effect that Meridian’s credit risk team noticed at once: government bonds are better collateral than equities for the indemnity. Stage 5.7 shows that this lowers Meridian’s capital.
The tri-party agent’s seat. For the tri-party team, the schedule is not a policy but a set of daily obligations. Every morning the engine revalues the collateral, applies the haircuts and the ESG flags, checks concentration limits, and allocates or substitutes collateral before the opening of the settlement day. A security that becomes ineligible overnight, because a vendor reclassifies an issuer or because Aurora updates its list, must leave the account within the grace period written in the schedule (Meridian proposes two business days). Three costs follow for Meridian as tri-party agent: the data licence, which grows with the number of clients using the same screens; the coding and testing of each new schedule; and the exceptions, when a borrower runs short of eligible collateral and the loan has to be partly returned. These costs explain why Meridian charges an ESG service fee rather than absorbing them in the split. A good tri-party agent also tells the client what the screen does not do: it controls what the client would hold after a default, and nothing else. It does not change the borrower’s own holdings or activities.
WHY IT MATTERS ON THE DESK — On the tri-party desk, an ESG schedule is judged by three numbers: the share of borrowers’ inventory that fails, the share of days with exceptions, and the number of breaches the client finds before Meridian does. The third should be zero.
Question. What is the expected and the tail loss of the borrower-default indemnity on the equity loans against non-cash collateral?
Data (case assumptions). Loans of EUR 630m at a 105% margin. Annual volatility of the lent equities 25%; of the collateral pool (half bonds, half equities) 14%; correlation 0.6. If a borrower defaults, Meridian needs 10 business days to sell the collateral and buy back the lent securities. Borrowers are large broker-dealers and banks rated A (PD 0.07%). The largest borrower accounts for 25% of these loans. Meridian’s agency revenue is EUR 0.641m (stage 5.2).
FORMULA
and are the returns of the lent securities and of the collateral over the close-out period; the margin (1.05); the change in the gap between what Meridian must buy back and what it can sell, per euro lent; and the annual volatilities; their correlation; the close-out period in business days; the expected shortfall per euro lent if the borrower defaults; the over-collateralisation cushion; and the standard normal density and distribution functions. The expected loss is . A tail loss uses the 99.9% quantile: for the largest borrower.
INTERACTIVE LAB — The borrower-default indemnity: expected loss and tail loss
Estimate the shortfall the agent lender pays when a borrower defaults, from the margin, the volatilities, the correlation and the close-out period, and compare the tail loss with the agent’s annual revenue.
| Parameter you can move | Range | Default |
|---|---|---|
| Indemnified loans | 10 to 5000 EUR m | 630 EUR m |
| Collateral margin | 100 to 115 % | 105 % |
| Volatility of lent securities | 1 to 60 % p.a. | 25 % p.a. |
| Volatility of collateral | 0 to 60 % p.a. | 14 % p.a. |
| Correlation lent securities / collateral | -50 to 100 % | 60 % |
| Close-out period | 1 to 30 business days | 10 business days |
| Borrower PD (one year) | 0.03 to 5 % | 0.07 % |
| Largest borrower’s share of loans | 5 to 100 % | 25 % |
| Agent’s annual revenue on the programme | 0.05 to 10 EUR m | 0.64068 EUR m |
| Result at default settings | Value |
|---|---|
| Volatility of the gap over the close-out period | 3.985 % |
| Expected shortfall if a borrower defaults | 0.1995 % of loan |
| Expected loss a year | 880 EUR |
| Loans to the largest borrower | 157.5 EUR m |
| 99.9% shortfall on the largest borrower | 11.52 EUR m |
| Tail loss in years of agent revenue | 18.0 years |
Try it. Halve the close-out period to 5 days: the expected shortfall falls by about two thirds. Then set the correlation to zero, as when a borrower posts bonds against equities in a crisis: the shortfall rises sharply. Margin, speed of close-out and correlation are the three levers of indemnity risk.
Result. Over a 10-day close-out, the gap between lent securities and collateral has a volatility of 3.98% of the loan. The expected shortfall if a borrower defaults is 0.20% of the loan; with a PD of 0.07%, the expected loss on the EUR 630m is about EUR 880 a year. The 99.9% shortfall on the largest borrower (EUR 157.5m) is EUR 11.5m, about 18 years of Meridian’s agency revenue. Applying the same model to the EUR 210m of equity loans against cash (margin 102%, collateral volatility zero) gives an expected shortfall of 1.14% of the loan if default.
Reading. The indemnity is a deep out-of-the-money put written on the joint event “a large borrower defaults and markets move against the collateral during the close-out”. Its expected cost is negligible; its tail is many years of revenue. This shape (tiny expected loss, large tail) is why a lending RAROC based on expected loss alone would be meaningless, and why regulation charges capital. It is also why the indemnity carries wrong-way risk: borrowers default in stressed markets, which is when volatilities and close-out periods are longest. Meridian’s credit team therefore manages the indemnity with four levers rather than with pricing alone: borrower limits (no borrower above a fixed share of indemnified loans); margins by collateral pair, higher when lent equities are collateralised by bonds or by equities from another market, because correlation is lower; exclusion of wrong-way collateral (a borrower’s own shares, or shares of its group); and a tested close-out procedure, with a list of brokers and a playbook for selling collateral and buying in securities in a stressed week. The ESG screen helps marginally on the third lever, because it pushes equity collateral towards government bonds; it also lowers correlation between lent and posted assets, which raises the shortfall per euro of equity collateral replaced. In this case the first effect dominates for capital (stage 5.7) and both are small for expected loss. Lehman’s default in September 2008 is the case study every agent lender’s credit team knows: agents that had to close out Lehman loans did so in the most volatile weeks in decades.
Question. Does Meridian’s 20% share pay for the capital it uses, and how does the ESG package change the answer?
Data (Meridian conventions and case assumptions). Cost-income ratio 60% on fee revenue (course convention). Operational-risk capital allocated at Meridian’s marginal BIC coefficient of 15% (see D02 and E03): EUR 0.24375 of capital per euro of revenue. Counterparty capital for the indemnity: exposures computed with the CRR3 financial collateral comprehensive method, supervisory volatility adjustments of 20% for main-index equities (lent or received) and 2% for government bonds, risk-weighted with the F-IRB formula for financial institutions (PD 0.07%, LGD 45% for senior exposures to financial sector entities, maturity 0.5 year for securities financing, correlation multiplied by 1.25 for large regulated financial entities). Expected loss: the model of stage 5.6 applied to the three segments. Meridian’s credit risk function recognises the master netting agreements with each borrower; on this book it estimates that netting reduces exposure by 60% (case assumption). The ESG package: revenue lost to screens (tiers 1 and 2) and to the stewardship overlay, EUR 0.316m gross; an ESG service fee of EUR 60,000 a year charged to Aurora for screening, recall management and reporting; a data cost of EUR 40,000 a year (vendor licence and the tri-party team’s time); EUR 59.5m of equity collateral replaced by government bonds.
FORMULA
is the exposure after credit risk mitigation; the value of the securities lent; the volatility adjustment on the lent securities; the value of collateral received; the volatility adjustment on the collateral (a currency mismatch would add 8%).
Applied to the three segments: equity loans against non-cash, = EUR 167.3m; equity loans against cash, = EUR 37.8m; bond loans, = EUR 30.6m. Total EUR 235.7m.
INTERACTIVE LAB — Meridian’s RAROC on Aurora’s programme: fee view, indemnity view, and the split that clears the hurdle
Compute the agent lender’s RAROC with operational-risk capital only and with the counterparty capital of the indemnity, with or without the ESG package, and find the agent share of revenue that reaches the 12% hurdle.
| Parameter you can move | Range | Default |
|---|---|---|
| Gross programme revenue before ESG package | 0.5 to 30 EUR m | 3.2034 EUR m |
| Agent share of revenue | 5 to 50 % | 20 % |
| Cost-income ratio on lending revenue | 30 to 80 % | 60 % |
| ESG package applied (0 = no, 1 = yes) | 0 / 1 | 0 |
| Gross revenue lost to screens and stewardship | 0 to 5 EUR m | 0.316259 EUR m |
| ESG service fee | 0 to 1 EUR m | 0.06 EUR m |
| ESG data and team cost | 0 to 1 EUR m | 0.04 EUR m |
| Equity collateral replaced by government bonds | 0 to 330 EUR m | 59.535 EUR m |
| Exposure reduction from netting | 0 to 90 % | 60 % |
| Allocated BIC coefficient | 12 to 18 % | 15 % |
| Result at default settings | Value |
|---|---|
| Agent revenue | 0.6407 EUR m |
| Agent’s share of revenue lost to the ESG package | 0.0000 EUR m |
| Operating costs | 0.3844 EUR m |
| Expected loss on the indemnity | 0.00256 EUR m |
| Exposure after collateral: equity loans, non-cash | 167.265 EUR m |
| Exposure after collateral: equity loans, cash | 37.800 EUR m |
| Exposure after collateral: bond loans | 30.600 EUR m |
| Exposure after collateral, before netting | 235.67 EUR m |
| Counterparty capital before netting | 4.6569 EUR m |
| F-IRB risk weight | 15.20 % |
| F-IRB risk weight without the 1.25 correlation multiplier | 11.21 % |
| Counterparty capital after netting | 1.8628 EUR m |
| Operational-risk capital | 0.1562 EUR m |
| Pre-tax profit after expected loss | 0.2537 EUR m |
| After-tax profit | 0.1903 EUR m |
| Profit after a 12% charge on allocated capital (netting view) | -0.0520 EUR m |
| RAROC, fee view (no indemnity) | 123.1 % |
| RAROC, indemnity without netting | 3.95 % |
| RAROC, indemnity with netting | 9.42 % |
| Agent share that reaches 12% (with netting) | 25.99 % |
Try it. Set netting to 0%: the RAROC collapses to about 4%. Then set the agent share to 30% with netting at 60%: the programme clears the hurdle. The answer depends far more on how the indemnity is capitalised than on the ESG package.
Result, baseline (no ESG package, 80/20). Meridian’s revenue is EUR 0.641m, operating costs EUR 0.384m, expected loss on the indemnity about EUR 2,600 a year. Three RAROCs:
Result, with the ESG package. Meridian loses EUR 63,000 of revenue through the split, earns the EUR 60,000 fee and pays EUR 40,000 of data cost; but the collateral shift to government bonds reduces exposure to EUR 224.9m. RAROC with netting rises slightly, to 9.64%. At a 75/25 split with the ESG package, it reaches 11.66%.
Reading. Four lessons. First, the ESG package is roughly neutral for Meridian: the fee covers the data cost and the agent’s share of lost revenue, and better collateral saves capital. Its real cost falls on Aurora (stage 5.4 and 5.5). Second, the programme as priced does not earn its counterparty capital, and the gap is not small: Meridian would need 26% of revenue, or a separate indemnity fee, to clear the hurdle on a standalone basis. Third, the answer is dominated by a modelling choice, netting recognition, which moves the RAROC from 4% to 9%; a committee should ask how robust that 60% is. Fourth, the fee view (123%) is not wrong, it answers a different question: what the business earns on its operational risk if the client carries the credit risk.
The decision. The new business committee offers Aurora two options. Option A: indemnified programme at 75/25 with the ESG package, RAROC 11.7%, accepted below hurdle because Aurora’s custody mandate and FX flows sit in the same relationship. Option B: 85/15 without indemnity, Aurora’s own credit team approving borrowers, with a fee-view RAROC far above the hurdle. Aurora, which has a credit team for its bond portfolio, takes a week to decide; this is the conversation that a trustee board should have, and rarely does, about what the indemnity is worth.
| Item | Value | Stage |
|---|---|---|
| Gross programme revenue | EUR 3.203m | 5.2 |
| Aurora / Meridian (80/20) | EUR 2.563m / EUR 0.641m | 5.2 |
| Aurora’s revenue on lendable assets | 2.1 bp | 5.2 |
| Cash collateral pool | EUR 214.2m | 5.2 |
| Extra income of enhanced reinvestment (Aurora) | EUR 0.428m a year | 5.3 |
| Stress loss of enhanced reinvestment (2%) | EUR 4.284m, 10 years of extra income | 5.3 |
| ESG filter on reinvestment | EUR 42,840 a year | 5.3 |
| Recall policy for flagged meetings | EUR 43,910 a year | 5.4 |
| Borrower restrictions | EUR 120,204 a year | 5.4 |
| Never lending voted shares | EUR 801,360 a year | 5.4 |
| Screens tiers 1-2 / tiers 1-3 | EUR 152,145 / EUR 375,406 a year | 5.5 |
| Collateral moved to government bonds | EUR 59.5m | 5.5 |
| Indemnity: expected shortfall if default | 0.20% of loan | 5.6 |
| Indemnity: expected loss (equity, non-cash) | about EUR 880 a year | 5.6 |
| Indemnity: 99.9% loss, largest borrower | EUR 11.5m, 18 years of agent revenue | 5.6 |
| Exposure after collateral | EUR 235.7m (EUR 224.9m with ESG) | 5.7 |
| F-IRB risk weight (financial institution, A) | 15.2% | 5.7 |
| RAROC: fee view / no netting / netting | 123% / 3.95% / 9.42% | 5.7 |
| RAROC with ESG package, 80/20 / 75/25 | 9.64% / 11.66% | 5.7 |
| Agent share for 12% (netting, no ESG) | 26.0% | 5.7 |
IN ONE SENTENCE — Agency lending without indemnity consumes only operational-risk capital, proportional to revenue; with an indemnity it also consumes counterparty capital, proportional to exposure, and the second can be ten times the first.
Operational risk. Since 1 January 2025 EU banks compute operational-risk own funds with the single standardised approach of CRR3. Fee and commission income enters the services component of the business indicator, and the business indicator component applies marginal coefficients of 12%, 15% and 18% by size; the EU sets the internal loss multiplier to 1 (see D01 and D02). Meridian allocates this capital to business lines at its marginal coefficient of 15% (see E03), so each euro of agency revenue carries about 24 cents of capital. Nothing in that charge depends on the indemnity, on the quality of the collateral or on the controls.
Counterparty credit risk of the indemnity. The indemnity makes Meridian economically exposed to each borrower for the difference between what it would have to buy back and the collateral it could sell. Supervisors treat it accordingly: the agent computes an exposure as if it were the lender, for the guaranteed difference. Three building blocks give the capital:
Three further rules matter. The output floor, phased in from 50% in 2025 to 72.5% in 2030, compares the bank’s total RWA with the standardised approach; because standardised weights for bank counterparties are higher than this F-IRB weight, the floor can make the indemnity more expensive at group level. Large exposures: indemnified loans count towards the exposure to each borrower group, limited to 25% of Tier 1 capital. Leverage ratio: for an agent whose indemnity is limited to the difference between the securities and the collateral, the Basel leverage framework counts only the current exposure (securities minus collateral, floored at zero) in the exposure measure, not the full value lent; with daily margining at 102% to 105%, this is usually small.
Meridian’s convention is . Line by line, for the baseline programme (stage 5.7, default values):
| Line | Value (EUR m) | Comment |
|---|---|---|
| Agent revenue (20% of EUR 3.203m) | 0.6407 | Fee split |
| Operating costs (60%) | 0.3844 | Course convention |
| Expected loss on the indemnity | 0.0026 | Stage 5.6 model, three segments |
| Pre-tax profit | 0.2537 | |
| After-tax profit (25%) | 0.1903 | |
| Operational-risk capital | 0.1562 | 24 cents per euro of revenue |
| Counterparty capital, no netting / netting | 4.657 / 1.863 | 13% x 15.2% x exposure |
| RAROC: fee view / no netting / netting | 123% / 3.95% / 9.42% | |
| Profit after a 12% capital charge (netting) | −0.0520 | Value destroyed a year |
Why the fee RAROC is a margin in disguise. In the fee view the numerator is and the denominator , where is revenue, the cost-income ratio and the capital per euro of revenue. Revenue cancels: , whatever the size of the programme. The ratio measures the operating margin, not the return on a risk taken. A lending RAROC is different in kind: the denominator is driven by exposure, PD, LGD and maturity, and the revenue (the margin over FTP) is the price of exactly that risk, negotiated loan by loan. Change the borrower’s rating and both numerator and denominator move.
The indemnity sits in between, and that is the problem. Its capital behaves like a lending exposure, but its revenue does not: Meridian’s 20% is a service fee negotiated against competing agents, not a price for the indemnity. Nothing in the fee rises when Meridian lends to a weaker borrower, posts more equity collateral or concentrates on one counterparty. That mismatch is why indemnified agency lending looks excellent in the fee view and mediocre in the indemnity view, and why some agents now price the indemnity separately, as an explicit fee or a different split by borrower type, as the stage 5.7 committee did.
What the sustainable feature changes. Little for Meridian, much for Aurora. The ESG package reduces gross revenue, of which Meridian bears its 20%; the ESG fee covers this and the data cost; the collateral shift lowers exposure. The net effect on Meridian’s RAROC is a small rise (9.42% to 9.64%). The cost falls on the client: about EUR 0.32m a year of gross revenue lost to screens and stewardship (stages 5.4 and 5.5), of which it bears 80%. The honest message to the client is that it is paying for policy consistency and voting rights, not for a lower-carbon programme.
An agency programme uses no funding: the securities and the collateral belong to the client and the borrower. FTP is therefore nil, with three exceptions. If cash collateral is left on deposit with Meridian rather than reinvested, it becomes a short-term wholesale deposit from a financial customer, with a high LCR outflow rate and little stable-funding value; Treasury prices it accordingly, usually below the overnight rate. If Meridian advances cash or securities to cover a borrower’s late return (to avoid a client’s failed sale), it creates an intraday or overnight exposure that Treasury charges for. And the tri-party business requires intraday liquidity for settlement of allocations and substitutions, which Meridian’s Treasury includes in its intraday liquidity framework. In a crisis, the indemnity itself can become a liquidity call: Meridian must buy back securities before it has sold the collateral.
For Aurora. Lent securities stay on Aurora’s balance sheet: although title transfers, Aurora keeps substantially all the risks and rewards (it receives manufactured income and gets equivalent securities back), so IFRS 9’s derecognition test is not met. Non-cash collateral received is not recognised unless it is sold or the borrower defaults. Cash collateral is recognised with a liability to return it; the reinvested assets are Aurora’s assets, measured under IFRS 9 like any other investment, which is why reinvestment losses hit Aurora’s accounts directly.
For Meridian. Nothing goes on the balance sheet. The fee share is revenue from contracts with customers under IFRS 15, recognised as the service is provided. The indemnity is an off-balance-sheet commitment: whether it falls within IFRS 9 as a financial guarantee contract or is provided for under IAS 37 depends on the bank’s accounting policy and on how the indemnified obligation is analysed; either way, the expected loss is small and the amounts indemnified are disclosed. US custodian banks, for instance, disclose the value of indemnified securities on loan in their annual reports.
Green Asset Ratio and BTAR. The GAR covers on-balance-sheet banking book exposures; agency lending has none. The separate key performance indicator for fees and commissions of credit institutions, which would have reached this business, was postponed by two years to 1 January 2028 by the simplification of the taxonomy disclosure rules (Commission Delegated Regulation (EU) 2026/73, adopted on 4 July 2025 and published on 8 January 2026).
Pillar 3 ESG. The ESG templates focus on banking book exposures by sector, maturity and collateral; the indemnity does not appear as a financed exposure.
PCAF. Agency lending neither finances an issuer (financed emissions) nor brings a new issue to market (facilitated emissions, see C01 and C05). Meridian therefore reports no emissions for this business, and should resist the temptation to report “ESG-screened collateral” as a sustainability achievement in its own reports.
What does matter for Meridian’s reporting is liability and conduct. The ESG options are part of Meridian’s product offering described in its sustainability statement, and claims made to clients (screens, recalls, reporting) are subject to the same greenwashing scrutiny as any other product claim. A schedule that fails to exclude what it promises to exclude is a conduct issue, not a carbon issue.
| Risk | Control | Owner |
|---|---|---|
| Borrower default with collateral shortfall | Borrower limits, margins by collateral pair, daily marking, close-out playbook | Credit risk; agency lending traders |
| Wrong-way collateral (borrower’s own shares) | Hard exclusion in tri-party schedule | Tri-party collateral management |
| Cash reinvestment loss | Conservative guidelines, maturity and liquidity limits, client sign-off | Cash reinvestment; client |
| Missed recall before record date | Recall calendar, deadline engine, fail escalation | Recall desk; proxy voting team |
| Screen not applied (data gap, coding error) | Default-to-ineligible rule, four-eyes coding, daily exception report | Tri-party collateral management; ESG data |
| Unfair allocation across lenders | Documented queue, compliance testing | Agency lending; compliance |
| Mis-selling of ESG features | Approved wording, review by compliance | Programme manager; compliance |
| Capital underpricing of the indemnity | RAROC review at onboarding and annually | Finance; new business committee |
| Legal enforceability of netting | Legal opinions per borrower jurisdiction | Legal |
Red-flag checklist for an ESG lending programme
Size and revenue (dated).
| Data point | Value | Source and date |
|---|---|---|
| Global lendable securities | EUR 37.4 trillion | ISLA market report, H1 2025 (end-June 2025) |
| Global securities on loan | EUR 3.0 trillion | Same |
| Global lending revenue, H1 2025 | EUR 5.6bn | Same, from S&P Global data |
| Global lending revenue, 2025 | Nearly USD 15bn, a record, up 22% | ISLA market report, 27 February 2026 |
| Non-cash collateral in European tri-party | Equities 46%, government bonds 38%, corporate bonds 14% | ISLA, H1 2025 |
| Euroclear collateral platform, daily average outstanding | EUR 2.1 trillion in 2025 | Euroclear 2025 results, 4 February 2026 |
What changed recently.
Open debates.
INTERACTIVE LAB — Check your understanding: agency lending and collateral management
Test the mechanics, the ESG overlay, the indemnity and the bank’s economics of agency lending.
Q1. Under the GMSLA, what does the lender lose while its shares are on loan?
Answer: C. Title passes to the borrower, who manufactures dividends back; the lender keeps price exposure and recall rights but cannot vote until the shares return.
Q2. What does a standard borrower-default indemnity cover?
Answer: A. The indemnity covers the collateral shortfall after a borrower default; reinvestment losses stay with the client.
Q3. In the case, Aurora’s lending income on its total lendable assets is about:
Answer: D. EUR 2.563m on EUR 12bn; 13.7 bp is gross revenue on value on loan.
Q4. Enhanced reinvestment earns Aurora EUR 0.428m a year more. A 2% stress loss on EUR 214.2m costs:
Answer: B. EUR 4.284m divided by EUR 0.428m is 10 years; the indemnity does not cover reinvestment.
Q5. Which stewardship option was cheapest in the case?
Answer: A. Recalls cost about EUR 44,000 against EUR 120,000 for restrictions and EUR 801,000 for withholding.
Q6. Why is the cost of ESG screens convex across tiers?
Answer: B. The funding differential of the replacement rises as cheap substitutes run out, and the rating floor excludes the largest share.
Q7. What did ISLA argue in its June 2022 paper on ESG collateral?
Answer: C. ISLA asked regulators to confirm that loans and collateral are not sustainable investments in themselves and that collateral need not be screened as strictly as a portfolio.
Q8. Which three parameters drive the indemnity’s expected shortfall in the stage 5.6 model?
Answer: D. The shortfall depends on the cushion (margin minus one) relative to the volatility of the gap over the close-out period.
Q9. Why is the F-IRB risk weight for the borrowers 15.2% rather than 11.2%?
Answer: B. CRR applies a 1.25 multiplier to the correlation for large regulated financial sector entities.
Q10. In the fee view, Meridian’s RAROC of about 123% is mostly a function of:
Answer: A. With capital proportional to revenue, revenue cancels and the ratio is (1 - c)(1 - t)/k.
Q11. What is the leverage exposure of an agent whose indemnity is limited to the difference between the securities and the collateral?
Answer: D. The Basel leverage framework counts only the guaranteed difference for such agents.
Q12. Why did the ESG package slightly raise Meridian’s RAROC in the case?
Answer: B. Exposure after collateral fell from EUR 235.7m to EUR 224.9m and the fee offset the agent’s share of lost revenue.
Q13. For Aurora’s accounts, lent securities:
Answer: C. IFRS 9 derecognition fails because Aurora keeps the economic exposure and gets equivalent securities back.
Q14. Which statement about sustainability metrics is correct?
Answer: C. Delegated Regulation (EU) 2026/73 postponed the fees and commissions and trading book KPIs by two years.
A foundation lends EUR 4,000m of equities at 15% utilisation and 30 bp (all against non-cash collateral at 105%, of which 40% equities) and EUR 3,000m of government bonds at 25% utilisation and 6 bp. The split is 85/15. (a) Compute gross revenue, the split and the client’s revenue on lendable assets. (b) Tiers 1 and 2 exclude 10% and 12% of equity collateral with funding differentials of 25 bp and 35 bp. Compute the annual cost, its share of gross revenue and who bears it.
Solution. (a) Equities on loan EUR 600m earn EUR 1.80m; bonds on loan EUR 750m earn EUR 0.45m; gross EUR 2.25m. Client EUR 1.9125m, agent EUR 0.3375m. On EUR 7,000m of lendable assets the client earns 2.73 bp. (b) Equity collateral subject to screens: = EUR 252m. Tier 1: = EUR 63,000; tier 2: = EUR 105,840. Total EUR 168,840, 7.5% of gross revenue; the client bears EUR 143,514 and the agent EUR 25,326. The screen costs more here than in the case because the collateral is mostly equities and the differentials are higher.
An agent indemnifies EUR 400m of equity loans against cash at 102%. Lent equities have a volatility of 30%; close-out takes 5 business days; borrower PD is 0.20%; the largest borrower has 30% of the loans. Compute the expected shortfall if default, the expected loss and the 99.9% loss on the largest borrower.
Solution. With cash collateral, = 4.23%; the cushion is = 2%; = 0.473. = 0.87% of the loan. Expected loss: = about EUR 7,000 a year. Tail: = about EUR 13.3m. The thin cushion of a cash-collateralised loan makes the shortfall per default four times larger than in the case’s non-cash book, even with a shorter close-out.
A programme earns EUR 2.5m gross; the agent keeps 20%. Cost-income 60%, expected loss EUR 5,000, exposure after collateral EUR 150m, risk weight 15%, netting reduces exposure by 50%, operational-risk capital EUR 0.24375 per euro of revenue, capital 13% of RWA, tax 25%. (a) Compute the RAROC in the fee view and with the indemnity. (b) What agent share reaches 12%?
Solution. (a) Revenue EUR 0.5m; costs EUR 0.3m; pre-tax after expected loss EUR 0.195m; after tax EUR 0.146m. Operational-risk capital EUR 0.122m; counterparty capital = EUR 1.463m. Fee view: = 123%. Indemnity view: = 9.2%. (b) With the agent’s revenue: , so = EUR 0.662m, a share of 26.5% of gross revenue. As in the case, about a quarter of the revenue is needed to pay for the indemnity’s capital.
1. Explain agency securities lending to a trustee in two minutes. The fund’s custodian lends its securities, for a fee, to banks and broker-dealers that need them, mostly to cover short sales or settle trades. The borrower posts collateral worth more than the securities, marked to market every day and held by a tri-party agent. The fund keeps the economic exposure through manufactured dividends and can recall the securities at any time. The agent keeps a share of the fee, typically 10% to 30%, and usually indemnifies the fund if a borrower defaults and the collateral falls short.
2. What does the indemnity cover and what does it not? It covers the shortfall between the cost of buying back the lent securities and the proceeds of the collateral after a borrower default. It does not cover losses on reinvested cash collateral, which follow the client’s guidelines, nor revenue lost to recalls or restrictions. The difference matters, because in 2008 the largest losses came from reinvestment. A trustee should read the scope clause of the authorisation agreement before relying on the word “indemnified”.
3. How would you price the ESG screens a client asks for in its collateral schedule? I would run each tier against the recent collateral inventory of the main borrowers to measure the share that fails. I would estimate the funding differential of the replacement collateral with the traders, which rises as cheap substitutes run out. The annual cost is the excluded collateral times that differential; borrowers pass it on through lower fees, so client and agent bear it pro rata to the split. I would show the client the cost by tier, and the data coverage, before it chooses.
4. Is lending compatible with responsible ownership? Yes, with conditions. The lender must be able to recall in time for the meetings that matter, which requires a calendar, a recall engine and reporting on recall effectiveness. It should know its borrowers and screen its collateral to stay consistent with its own exclusions. Withholding voted shares from lending costs several times more than recalling them, so the case for lending with recall is strong.
5. Why is the RAROC of a fee business different in nature from a lending RAROC? Under CRR3 operational-risk capital is proportional to revenue, so in a pure fee view revenue cancels and RAROC equals roughly the after-tax margin divided by the capital per euro of revenue; it measures cost efficiency, not risk-adjusted return. In lending, capital depends on exposure, PD, LGD and maturity, and the margin is the price of that risk. Agency lending with an indemnity mixes the two: capital like a loan, revenue like a fee. That is why it can look excellent in one view and below hurdle in another.
6. What drives the capital of an indemnified programme, and how can you reduce it? The exposure after collateral, under the comprehensive method, depends on the volatility adjustments of the lent securities and of the collateral; equity against equity is expensive, government bonds against government bonds cheap. The F-IRB weight of financial counterparties includes a higher correlation. Netting recognition under enforceable master agreements reduces exposure a lot. Levers: better collateral mix, netting opinions, borrower selection, and pricing the indemnity explicitly.
7. A client wants to reinvest cash collateral in one-year bank paper to raise income. What do you say? I would show the extra income against the loss in a stress: in the case, 25 bp of extra yield is wiped out ten times over by a 2% loss. The loss would be the client’s alone, because the indemnity does not cover reinvestment. Lending is meant to earn income on idle securities, not to run a leveraged credit fund. If the client insists, maturity, credit and liquidity limits must be tight and approved by its own investment committee.
8. Can a lending programme be called sustainable? Not on the basis of screened collateral or recalls. Collateral is held only if a borrower defaults, and the loan finances no green asset. What the programme can claim is consistency with the owner’s policies: votes exercised, counterparties known, residual holdings aligned with exclusions, cash reinvested within screened guidelines. That is worth reporting precisely, with its cost, rather than labelling.
Industry frameworks and data
Regulation and supervisory texts
Theory and textbooks