A Rs 2,400 crore credit risk fund faces Rs 600 crore of redemptions in a week with 9% in cash and treasury bills. Selling its most liquid bonds first would cut AAA holdings from 42% to about 23% of what remains. Sell the liquid assets, sell a vertical slice, or borrow? Who bears the cost in each?

Sell a vertical slice, so the fund that remains is the fund investors bought, and the cost of leaving is shared by those leaving. It realises about Rs 3.6 crore of haircuts, 15 basis points across everyone. Selling liquid assets first costs nothing today but leaves stayers with AAA at about 23% and liquid assets at 35% instead of 51%, and the next wave would cost them about 9 basis points. Borrowing only defers the sale at the stayers' expense.

Six flatmates split a fridge. Two move out and take all the fresh food because it is easiest to carry, leaving the other four with tins. Nobody paid anything, and the four are worse off. A fund meeting redemptions is choosing which investors bear the cost of turning bonds into cash, and the easiest choice, selling what sells, quietly hands that cost to whoever stays. The question is not which sale is cheapest today; it is which sale leaves the remaining investors with the fund they were sold and makes the leavers pay for their own exit.

It is painless on the day. Rs 600 crore of AAA bonds go at almost full price, about Rs 0.6 crore of haircut, and the cash buffer is kept because the fund's liquidity rules need it. But the fund that remains has AAA at about 23% instead of 42%, A and below at 25% instead of 19%, and assets it can sell quickly at 35% instead of 51%. On the credit team's own loss assumptions, expected losses rise from about 21 to about 27 basis points a year, which is modest; the real damage is the next week. If another Rs 600 crore leaves, the fund has only Rs 408 crore of AAA left and must sell AA at a 0.6% haircut, costing stayers about 9 basis points, and the fund after that is mostly A-rated paper that nobody is bidding for. Runs feed on exactly this: the investors who understand it leave first.

Sell a quarter of every holding. The haircuts on the AA and lower-rated paper are real: about Rs 3.6 crore in all, 15 basis points of the fund. Because the sales are made before the day's NAV is struck, that cost is in the NAV everyone receives, so the leavers bear their share of it and the stayers keep the fund they bought. The portfolio afterwards is 42% AAA and 9% cash, the same as before, and the next wave costs the same 20 basis points, not more. The limit is practical: a quarter of the A-rated paper may not find a buyer in a week at any sensible price, so a real slice is tilted towards what trades, with the tilt disclosed to the committee.

Only when the redemptions are a passing panic and the money is likely to come back. A month's borrowing of Rs 600 crore at 8% costs about Rs 4 crore, 22 basis points of the remaining fund, paid entirely by the stayers, while the leavers get full NAV today. If the redemptions continue, the fund has paid interest and still has to sell, so borrowing is a bridge, not a plan, and it is used to avoid dumping bonds into a one-week panic, not to avoid selling at all. The regulations cap how much a scheme can borrow and for how long; the committee should confirm the current limits before relying on it.

The truth, in numbers: a quarter of the fund left, the fund sold a quarter of everything, the haircut was shared, and the portfolio they own is the one they owned last week. The tools that make leavers pay their own way, such as exit loads, swing pricing that adjusts the NAV on days of large flows, and a standing liquid buffer, work only if they exist before the run, so the second recommendation to the committee is to check which of them this scheme has. The regulator has set rules on liquidity buffers and on swing pricing for some debt funds; confirm the current framework. The limit of this analysis is the desk's haircut estimates: in a real panic the A-rated paper may have no bid at all, and then the honest choice is between a tilted slice and, in the extreme, restricting redemptions, which the rules allow only in defined circumstances.