Adverse Selection in Motor Insurance Underwriting

In June 2026, ICICI Lombard traced a motor insurance fraud network in Karnataka: an FIR against nine people, a run of staged accident-compensation claims, and witnesses vouching for accidents they were never at. 

No single claim gave it away. The repetition did. 

The same names, the same method, across a large series of cases. Months earlier, the Supreme Court had flagged something similar as a ‘wide racket’ of fabricated accident claims.

Once fraud is organised and repeatable, it concentrates wherever a process is easiest to work around. In motor insurance, investigations sit at the claims stage, after the vehicle is already insured. And by the time it concludes, money is already gone and losses are locked in. 

Why the claims stage is a difficult place to control risk

A strong investigation team can flag a suspicious third-party claim correctly and still not change the outcome.

• Tribunals exist to compensate victims, and their awards bind. Even with some evidence of fraud, the insurer is often directed to pay.
• Investigation draws on the most experienced people in the organisation, for claims that may be paid regardless.
• A contested third-party claim can sit for five to ten years, reserves locked, legal costs accruing throughout.

The ₹5-8 lakh settlement often quoted currently is only the visible part of what a high-risk vehicle costs. Claims investigations, while necessary, manage losses which are already on the book rather than controlling whether they arrive or not.

IRDAI Regulations now point to the same limitation.

IRDAI's Insurance Fraud Monitoring Framework Guidelines, 2025, in force since 1 April 2026, requires a board-approved anti-fraud policy reviewed annually, a Fraud Monitoring Committee and an independent Fraud Monitoring Unit, FMR-1 reporting with CEO certification, frauds above ₹1 crore reported within 30 days, and data sharing through IIB. It also extends accountability to distribution channels, where motor insurance sells through dealerships, aggregators and OEM partners and the gate sits outside the insurer's own systems.

What the vehicle already told you

A vehicle that turns into a problem later has almost always been a problem before, with some other insurer, and the record is usually available before issuing it a policy. 

We pulled the histories on two cars that recently surfaced in fraudulent-claim reporting:

1. A Chevrolet Tavera, GJ 09 BA 1430, insured in mid-2025. Police had already seized it twice, in August and December 2024. Later reporting added a series of linked claims, the same individual appearing as both owner and driver, and the same names recurring over several years.

Months after the policy was written, the vehicle was seized again.

2. A Maruti Swift, MP 06 CA 2111, insured in mid-2025. By then it carried four prior incidents: a seizure in June 2018, a court case that December for causing death by rash and negligent driving, and two more seizures in 2022 and 2023.

The car later turned up in a fake-claim investigation with three registered owners and staged accidents.

In neither case was the risk especially hard to find. It was documented with FIRs, seizures, a court case and available in public records at the point of underwriting.

A history like this functions as a signal, and in both cases it predated the cover

But Industry data alone wouldn't have flagged it. 

The Insurance Information Bureau (IIB) runs a claim propensity score for the private car own-damage portfolio, positioned at the underwriting stage to green-channel low-score vehicles, and carries NCRB stolen-vehicle data alongside VAHAN.

What it doesn't hold is the adverse legal record. Seizures, FIRs and criminal cases sit with the police and the courts. 

These scattered data points, triangulated, show an insight into actual risk.  

From our work with insurers' claims and investigation teams, around 8-10% of private vehicles and 12-15% of commercial vehicles already on the books, carried an adverse legal history, bad enough to mark them high-risk.

Run the same check at underwriting and the share would be smaller. This drop is the point, not a flaw. At claims you only see cars that already misbehaved; at underwriting you see the whole intake. 

Risk clusters in a thin slice of vehicles carrying their past with them, and that slice needs to be visible at the gate.

Motor profitability, or the lack of it, usually gets labelled as a pricing problem or a fraud-detection problem. Whereas, a significant portion is actually a risk selection problem, visible in motor insurance underwriting before the policy is issued.

The economics of selection are fairly simple

A third-party claim usually costs ₹5-8 lakh. The average motor premium is about ₹5,000. One claim is worth roughly a hundred policies' worth of premium.

The trade-off is lopsided. 

Declining a small tail of vehicles with an adverse history forgoes a few thousand rupees each. One avoided claim offsets around a hundred of those rejected policies, not to mention the investigation costs, the reserves and the years of litigation avoided.

This isn't a case for declining more people.

Most vehicles currently prove to be clean and should pass straight through, and in a market this competitive, nothing else is viable. Risk selection works as a narrow filter aimed at the small group whose background and predicted risk tends to drive disproportionate losses.

This method isn't foolproof either.

Records vary in completeness across states, and a policy reject has to be defensible, which is why it should rest on a documented past rather than an opaque, indefensible guess. 

Where the balance is shifting to now

Motor underwriting has always been at least as much about selection as about pricing, and selection is a decision best made at the gate. With third-party rates administered and own-damage rates competing away, composition of the book is the only lever with room left in it.

Regulation is moving the same way. Detection at claims will stay necessary. The open question is how much of the effort moves upstream. In practice that means resolving the vehicle and parties to canonical identifiers at quote stage, pulling documented adverse history against them, triangulating rather than acting on one stale record, and tiering the outcome so the majority pass straight through and only a clear pattern triggers a decline. 

Risk orchestration layers such as IDfy's OneRisk run that sequence inside a live quote, in the seconds a motor journey allows.

The information already exists: the trail a repeat-offending vehicle leaves, the record it carries, the way tribunals weigh in favour of claimants. What an insurer chooses is when to consult it. 

At the claims stage the exposure is already on the book. At underwriting, there's still a decision about whether it should belong in the first place.

FAQs

What is underwriting in motor insurance? It's the process of assessing a vehicle and its owner to decide whether to offer cover, on what terms and at what premium. Because third-party rates in India are administered, motor underwriting turns less on pricing than on which vehicles an insurer accepts.

What is adverse selection in motor insurance? It occurs when higher-risk applicants are likelier to seek cover than lower-risk ones, leaving the insured pool worse than the general population. In motor, vehicles with a history of accidents, seizures or repeated claims cluster with the insurers applying the lightest checks at onboarding.

Can an insurer decline a motor policy based on a vehicle's legal history? Insurers can apply risk-based acceptance criteria for own-damage cover, though third-party cover is compulsory under the Motor Vehicles Act, 1988. A decline has to be defensible, which means resting it on a verifiable record rather than an unexplained score, and keeping the evidence trail.

How long does a contested third-party motor claim take to resolve? Five to ten years once appeals are counted, with reserves locked and legal costs accruing throughout. That's why the headline settlement figure understates what a high-risk vehicle actually costs.

What checks can be run on a vehicle before issuing a motor policy? Registration and ownership can be validated against VAHAN, while IIB supplies claim history, NCRB stolen-vehicle data and fraud triggers such as multiple policies across insurers. Adverse legal history sits outside insurance databases and needs a separate check against public records.

What does the IRDAI Insurance Fraud Monitoring Framework, 2025 require? Since 1 April 2026: a board-approved anti-fraud policy reviewed annually, a Fraud Monitoring Committee and independent Fraud Monitoring Unit, FMR-1 reporting with CEO certification, frauds above ₹1 crore reported within 30 days, and data sharing through IIB.