A dividend is declared and half the client's holding is out on loan while part of the rest is pledged as collateral. Who receives the income, who votes, and what does your system show the client?

One economic position, several legal ones

The client thinks they own a number of shares. Legally they own rather less than that, and the difference is the whole of this question. A securities loan is not a rental: title passes to the borrower, who becomes the owner of those shares and is obliged under the lending agreement to return equivalent securities and to make the lender economically whole in the meantime. A pledge may or may not transfer title depending on how the collateral arrangement is documented. So on the record date, the entity on the register for a given slice of the holding may be the borrower, the collateral taker, or your own nominee, and the entitlements follow the register.

Asset servicing is the business of reassembling one economic answer for the client out of those several legal positions. That is why the position model has to carry location and legal state, not just quantity: available in the depository, out on loan to a named borrower, delivered under a title-transfer collateral arrangement, held as security interest, pending settlement in either direction. A single quantity cannot answer any of the questions the client will ask.

Income on the lent portion is manufactured, not received

The borrower owns the lent shares on the record date, so the issuer pays the dividend to the borrower. Under the lending agreement the borrower then pays the lender an equivalent amount, a manufactured payment. Economically the lender is intended to end up where they would have been, and operationally it is a different payment, from a different party, on a different timetable, under a different legal character.

Three consequences fall out of that. The payment carries counterparty risk that the real dividend did not, which is why lending programmes are collateralised and why the collateral has to be adequate across a record date rather than on average. It can arrive later than the real dividend, so a client report that shows income as received on payment date will be wrong for the lent slice. And it can be taxed differently, because a manufactured payment is not a dividend and its treatment depends on the jurisdictions of both parties and on the agreement. The lender can therefore be worse off net than if they had never lent, even though the gross amounts match, and whether the fee earned covers that gap is exactly the calculation the lending desk should be doing before a record date. Where cross-border lending is being used specifically to change which jurisdiction's withholding rate applies to a dividend, tax authorities have taken a close interest, so this is a control question and not only an optimisation.

flowchart TD
    A[Dividend declared with ex and record dates] --> B[Record-date position split by location]
    B --> C[Available in depository]
    B --> D[Out on loan]
    B --> E[Pledged under title transfer]
    B --> F[Pledged as security interest]
    C --> G[Dividend received and allocated]
    D --> H[Manufactured payment from borrower]
    E --> H
    F --> G

The two pledge branches are the point: the same word covers two arrangements that send the entitlement to opposite places.

The pledge depends on the documentation, and you must not guess

If the collateral was delivered under a title-transfer arrangement, the taker owns it and the position behaves like a loan, with an obligation to return equivalent securities and to pass the economics back. If it was given as a security interest, the client keeps title and the taker has a right to enforce against it, so the dividend and the vote remain the client's. The system therefore has to carry the legal character of each collateral arrangement as a first-class attribute of the position, sourced from the agreement rather than inferred from the movement. This is the sort of detail that is captured correctly at onboarding and then lost the first time somebody builds a position report from settled quantities alone.

Voting is decided before the record date, and it is a trade-off

Only the record-date holder can vote. So the lent portion cannot be voted by the client, and the only way to recover the vote is to recall the loan early enough that the return settles before the record date. Working backwards, that means the recall notice period, the settlement cycle and any buffer for a failed return all have to fit between the decision and the record date, and the deadline that matters is therefore several days ahead of the record date rather than anywhere near the meeting.

That makes voting a commercial decision rather than an administrative one, and it belongs to the beneficial owner: recall the stock and lose the lending revenue, or keep the revenue and forgo the vote. Some owners have a policy that certain resolutions always justify a recall. Whichever way it goes, the system has to make it a decision somebody takes with the numbers in front of them, at a deadline it calculates, rather than a consequence the client discovers afterwards.

There is a further constraint on the custodian's side. Positions are usually held in omnibus accounts where many clients' holdings sit under one nominee, and the nominee cannot submit more votes than the record-date position. Over-voting is a real operational failure with real consequences, and the defence is that voting entitlement is allocated from the record-date position by client, with the lent and transferred slices excluded, before any instruction is accepted. A voting system that accepts instructions against traded quantities will over-vote the moment lending activity moves.

Deadlines stack, and the earliest one is yours

Every entitlement in this area has a chain of deadlines, and the market's deadline is the last of them. The registrar or the issuer's agent sets a date, the depository sets an earlier one, the global custodian sets an earlier one again, and your own operations team needs time before that. A client told the market deadline will miss it. Publishing the internal deadline, and holding the calculation of it as data per market and per event type rather than as knowledge held by an operations team, is the difference between an asset-servicing platform and a mailbox.

What the client should see

The report should lead with the economic position and the income they are entitled to, because that is what they own and what they are measured on. Underneath it, the legal breakdown has to be available: how much is available, how much is lent and to whom, how much is under each collateral arrangement, what is unsettled in each direction, and for each slice where the income is coming from and when. Accruals should be raised on the ex-date against the economic position, so that recalling a loan changes the source of the income rather than its amount, and the difference between accrued and received is a reconciliation item that ages rather than a figure that silently changes.

The client owns one economic position and the register knows several legal ones, and every entitlement follows the register. Custody systems that carry only a quantity cannot say who is going to be paid or who is allowed to vote, which are the only two questions anybody asks on a record date.