A bank at its Rs 500 crore limit to one borrower can buy five-year CDS protection on Rs 100 crore at 220 bp and lend Rs 100 crore more at a margin of 280 bp. What does the trade earn, and which risks does the bank still carry?

The trade nets 60 bp a year, about Rs 60 lakh on Rs 100 crore and Rs 3 crore over five years before discounting. In return the bank keeps the risk that Kumbharli Bank fails alongside Ratnagad, a recovery basis between the loan and the CDS payout, any loan life beyond five years, and profit swings because the CDS is marked to market while the loan is not. The Ratnagad limit is freed only by moving Rs 100 crore onto Kumbharli.

Picture a landlord who rents a flat to a tenant at Rs 28,000 a month and pays Rs 22,000 for rent-default insurance. His income is the difference, Rs 6,000, and his remaining worry is whether the insurer pays. Sonkhed Bank earns 280 bp on the loan and pays 220 bp for protection, so it keeps 60 bp a year, Rs 60 lakh on Rs 100 crore, as payment for every risk the CDS does not transfer. Over five years that is Rs 3 crore before discounting and before the cost of the capital the bank still has to hold.

It is worth reading what the 220 bp says about Ratnagad. If the market assumes 40% recovery, a 220 bp premium implies a default intensity of about 3.7% a year, which compounds to roughly a 17% chance of default within five years. The bank is lending at a margin that pays only 60 bp more than the market's price for that default risk, so the trade is a thin carry, not a cheap loan. The question the committee should ask is whether 60 bp pays for what is left.

A CDS settles on an auction of Ratnagad's deliverable debt, usually its bonds. The loan recovers whatever the bank collects from its security and the bankruptcy process. Those are two different numbers. If the loan recovers 60 and the auction prices the bonds at 40, the bank receives 60 from the CDS and 60 from the loan, Rs 20 crore more than par; if the loan recovers 30 and the auction prints 45, the bank gets 55 plus 30 and is Rs 15 crore short. A secured bank loan usually recovers more than unsecured bonds, so the basis often runs in the bank's favour, but it is not promised, and the definition of a credit event in the contract can also differ from the event that hurts the loan, for example a restructuring of the loan alone.

Because the protection pays only if Kumbharli Bank is solvent on the day Ratnagad is not. If both fail together, the bank recovers only the loan's 40 and has an unsecured claim on a failed bank for the rest, a loss of about Rs 60 crore, the same as if it had never bought protection. That joint failure is not a remote coincidence when the seller is another Indian lender with its own loans to the auto sector; it is wrong-way riskExposure to a counterparty that tends to grow, or a counterparty that tends to weaken, exactly when the underlying risk goes bad.. The committee should therefore count the Rs 100 crore against Kumbharli's limit, ask for collateral under the CDS, and prefer a seller whose business does not move with Indian autos.

Two quieter things. The CDS is revalued every day while the loan is held at cost, so reported profit moves even when nothing has happened to the cash flows. With a risky annuity of about 3.82 years, a widening of Ratnagad's spread to 400 bp shows a mark-to-market gain of about Rs 6.9 crore on the CDS, and a tightening to 120 bp a loss of about Rs 3.8 crore, with no offset on the loan; whether hedge accounting can pair them is a question for the bank's auditors under current rules. And the loan may be longer than the protection: if the plant loan amortises over seven years, the last two are unhedged. Whether the protection actually frees the regulatory exposure limit also depends on the regulator's credit risk mitigation rules, which the desk must confirm in their current form before relying on them.

So the honest view is narrow. The trade earns a real but thin 60 bp, and it is sound only if Kumbharli Bank is a strong, collateralised seller with little auto exposure, the loan's maturity and credit event definitions match the CDS, and the committee is content to swap a Ratnagad limit problem for a Kumbharli one. If any of those fails, the bank is being paid 60 bp to hold most of the risk it set out to remove.