Commission on Cancellations, Short-Period Policies and Refunds

When a policy ends before its expiry date, three numbers fall out in sequence: the premium the insurer keeps, the premium that returns to the insured, and the brokerage earned on the premium kept. Only the first is a real computation. The other two are consequences of it.

Most broking teams work the problem backwards. They take the unexpired days, compute a fraction of the annual premium, and treat the answer as the refund. That is right perhaps half the time, and the half where it is wrong is the half where the client calls.

A companion post on controlling commission reversals covers the control side: how reversals are detected, reserved against and reconciled once they exist. This post sits upstream, on what the number should be.

The idea that organises everything below: premium earned by an insurer is not a function of time. It is a function of exposure, and exposure is not spread evenly across a policy year. Every refund rule in the market prices that unevenness.

Indian general insurance uses two refund bases, and the selector is not the reason for cancellation. It is who ended the contract.

Pro-rata returns premium in strict proportion to the unexpired period. A twelve-month policy cancelled at day 200 returns 165/365 of the premium. It treats risk as uniformly distributed across the year, which is a fiction, but a fair one when the insured did not choose the ending.

Short rate retains more than elapsed time would suggest. Acquisition cost, survey, issuance and the reinsurance cession are all incurred at inception. More importantly, the insured who cancels early is not a random draw from the book: one who has had a clean six months and leaves differs systematically from one picked at random, and the scale prices that selection.

The pattern across most Indian wordings:

1. Insured cancels, insurable interest intact. Short rate. The insured elected to leave a bargain priced for a full year.
2. Insured cancels, insurable interest gone. Pro-rata. The asset was sold, the business closed, the project finished. There is no election to price.
3. Insurer cancels. Pro-rata, almost universally.
4. Void from inception. Full return, no scale, no apportionment.

The scale most Indian wordings reproduce descends from the tariff era. It expresses the premium charged for a short run as a percentage of the annual premium, not the percentage refunded. That single reading error is the most common mistake in the market.

The classic shape:

• Not exceeding 15 days: 10% of annual premium
• Not exceeding 1 month: 15%
• Not exceeding 2 months: 30%
• Not exceeding 3 months: 40%
• Not exceeding 4 months: 50%
• Not exceeding 5 months: 60%
• Not exceeding 6 months: 70%
• Not exceeding 7 months: 75%
• Not exceeding 8 months: 80%
• Not exceeding 9 months: 85%
• Exceeding 9 months: 100%

Scales vary by insurer and by wording, so the operative one is attached to the policy rather than held in anyone's memory.

Why day 62 costs eighty thousand rupees

The scale is banded, not continuous. Pro-rata is a straight line; short rate is a staircase. Take a fire policy with an annual premium of INR 8,00,000.

Cancelled on day 59, the run falls in the "not exceeding 2 months" band. The insurer charges 30 percent, INR 2,40,000, and refunds INR 5,60,000. Cancelled on day 62, it falls into "not exceeding 3 months": 40 percent, INR 3,20,000 charged, INR 4,80,000 refunded.

Three days of cover cost the client INR 80,000, a full 10 percent of the annual premium. Nothing in the risk changed. The client crossed a band boundary, and a broker who knows the scale can say so before the date is fixed.

At four months of run, short rate charges 50 percent and refunds INR 4,00,000, where pro-rata on 122 elapsed days would refund INR 5,32,603. The scale is worth INR 1,32,603 to the insurer on one mid-sized policy.

The scale has a second life most broking teams never meet, because they only encounter it at cancellation. Some policies are written on it deliberately, from inception, as the rating basis.

A fire policy on seasonal stock for the three months around a harvest, a burglary extension for a four-month exhibition, a top-up period bridging a client's programme to a new common renewal date: none of these is a twelve-month contract that failed. Each is a short-period contract priced by the same table, used forwards rather than backwards. The insurer charges 40 percent of the annual rate for the three-month exposure and earns all of it.

This matters twice over. First, brokerage on a deliberate short-period policy is earned in full at inception. Teams that treat any sub-annual policy as an anomaly and reserve against it are provisioning against revenue that will never reverse.

Second, you cannot apply the scale twice. Suppose a policy is written for five months at 60 percent of an annual rate of INR 20,00,000, so the premium charged is INR 12,00,000. The insured cancels after seven weeks. The refund is not a pro-rata slice of the INR 12,00,000, and it is not 30 percent of it either. The correct computation runs the scale against the annual rate for the period actually at risk: seven weeks falls in the "not exceeding 2 months" band, so the insurer retains 30 percent of INR 20,00,000, or INR 6,00,000, and refunds the balance of INR 6,00,000.

Naive pro-rata of the paid premium across the five-month term would have refunded INR 8,13,158. The gap of just over INR 2 lakh exists because two people applied a table to two different bases.

Engineering project policies (CAR, EAR) do not work this way. They are rated on the construction period as the policy term, with their own period scales and separately drafted early-completion terms.

Most Indian wordings carry a clause to the effect that no refund of premium is allowed where a claim has been made or paid. Clients read this as punitive. It is not. It follows from how an indemnity contract is built.

The premium buys a sum insured for a period, and that limit is single and indivisible, not a series of monthly allowances. An insured who suffers a INR 40 lakh fire loss in month three on a INR 5 crore sum insured has drawn on the whole contract, not on a quarter of it. The insurer has performed. There is no unearned portion to give back, because what the premium purchased was consumed the moment the limit was exposed.

The same logic explains outcomes that look inconsistent from outside.

1. Reinstated sum insured. Where the policy reinstates against additional premium, that reinstatement premium is separately earned and is not refundable either. The client bought a second limit.
2. Liability with an aggregate. A claims-made policy whose aggregate has been eroded has no unearned premium, and cancelling it leaves the insured exposed for acts already committed unless run-off is bought.
3. A notified but unsettled incident. Some wordings deny refund on any claim made, others only on a claim paid. That single word decides a real number.
4. Marine open covers. Premium attaches per declared shipment. There is no unearned annual premium to refund; the cover simply stops accepting declarations.

For brokerage the arithmetic is clean: no premium refund means no premium reversal, so the brokerage stands in full. A cancellation after a paid claim is one of the few endings that costs the broker nothing.

Two retail-adjacent lines that broking firms carry in volume have refund mechanics prescribed outside the wording, and both break the rules above.

Motor

Section 146 of the Motor Vehicles Act, 1988 makes third-party cover compulsory for any vehicle used in a public place. That statute, not the policy, governs what can be cancelled. The insurer requires evidence that the statutory gap does not arise: proof of a fresh policy for the balance period, or evidence the vehicle has been sold, scrapped or taken off the road.

Bundled structures make the arithmetic two-part. A new private car is sold with three years of third-party cover and one year of own-damage; a new two-wheeler with five years of third-party. Own-damage follows the wording's cancellation basis. The unexpired long-term third-party portion is a separate premium for a separate statutory period. A single blended fraction across a bundled motor premium is arithmetically meaningless.

On sale of the vehicle, nothing is refunded, because nothing is cancelled. Third-party liability follows the vehicle to the transferee by operation of law, and the own-damage section transfers by endorsement on application within 14 days of the transfer date. No-claim bonus stays with the seller, not the vehicle. Processing a vehicle sale as a cancellation destroys the client's NCB and opens a cover gap in one action.

Health

Retail health has moved away from short rate. Under the IRDAI Master Circular on Health Insurance issued in May 2024, cancellation at the policyholder's instance carries a pro-rata refund for the unexpired period rather than a retention scale. So one instruction, across two policies in the same client file, produces a staircase on the fire cover and a straight line on the health one.

Section 64VB of the Insurance Act, 1938 provides that no insurer shall assume risk unless the premium has been received in advance or guaranteed in the prescribed manner. It produces the one ending on this list that is not a refund at all.

Where premium was never received, or was tendered by an instrument that dishonoured, risk never attached. There is no policy period to apportion. The money in the insurer's hands never became premium; it was a receipt against a contract that failed to form. So:

• No scale applies. Not short rate, not pro-rata. The staircase and the straight line both need a run of cover to measure, and there was none.
• The return is of the whole sum actually received, without retention for the days the client believed they were covered.
• No brokerage was ever earned. Brokerage on a 64VB failure is not revenue that reverses, it is revenue that never existed. If it was booked, it was booked in error, and if it was paid to a producer, the firm paid an incentive on nothing.

A 64VB failure should therefore never be run through cancellation logic. The scale produces a number wrong by construction, and implies downstream that cover was in force for some period. It was not, and the notice to the client has to say so.

Once the retained premium is settled, the remaining numbers are mechanical. They are also asymmetric in time, and the asymmetry always runs one way.

1. The cancellation endorsement is issued, and its effective date is the fact everything else depends on. It should be the date cover actually ceased, not the date the instruction arrived and not the date the back office got to it. A cancellation processed three weeks late and back-dated crosses a band on the scale often enough to matter.
2. The insurer computes retained premium on the basis in the wording and refunds the balance.
3. The insurer issues a credit note under Section 34 of the CGST Act, 2017 for the refunded value plus tax, and the insured reverses the corresponding input tax credit.
4. The insurer adjusts the brokerage on the refunded premium, usually as a debit against a later month's commission statement.

Step four is where the timing sits. The insured is made whole in days; the broker's adjustment lands 30 to 90 days later, against a statement for unrelated business. In between, the insurer has refunded gross while still holding the broker's earlier deduction, and the broker funds the gap out of working capital. This is structural: brokerage is netted at inception and returned in arrears.

One tax wrinkle is worth flagging. Since individual life and health premium became GST-exempt, a cancelled retail health policy has no output tax on the premium to reverse, so there is no premium credit note and the refund is simply the money. Commercial lines still carry 18 percent: a cancelled fire policy at INR 8,00,000 refunding half generates a credit note for INR 4,00,000 plus INR 72,000 of tax, while the broker's own credit note on brokerage at 12.5 percent covers INR 50,000 plus INR 9,000.