The Carrier Name Is Not the Risk Bearer

The name on the broker portal is doing a job. It helps somebody recognise a market and decide whether to quote. It is not telling them who has promised to pay a claim.

The logo on the quote is marketing data. The legal entity on the declarations is the promise.

That is the first correction. The second is less comfortable. Even the legal insurer on the declarations may not retain most of the economic risk. In a fronting arrangement it issues the policy on its licence and cedes most or all of the risk onward. The issuer stays the legal carrier, and keeps credit risk if its reinsurer does not pay. The economic story has already moved elsewhere. Which is why I would not ask a system for “the carrier” and accept one string in reply. There are three questions. Who holds the licence. Who is named as obligor on the policy. And who bears the loss, net of reinsurance.

They commonly point to one company.

The cases that matter are where they do not.

A system that cannot keep those questions apart also can’t explain its own security when a broker asks, a complaint arrives, or a rating moves.

Not a finance-team enrichment exercise.

Issuance data, counterparty data, operating-control data.

The front end is allowed to be simple

A managing general agent can quote, bind, issue documents, price within guidelines, and often adjust claims. Its name may be the only one customers see.

Neither makes it the carrier.

An MGA holds delegated authority.

It holds no capital against the book, files no statutory returns, and is not the obligor.

Administrative services only makes the mistake look friendlier.

A life insurer can print the card, administer the plan, and run the network while the employer funds the claims. The name on the card is a servicing identifier.

Not evidence that the member has insured benefits.

A service-contract administrator and the insurer behind it repeat the confusion.

Nothing wrong with a simple distribution experience.

Something wrong with making its identity model our policy model.

The record we keep below the brand is more severe.

Per policy term: issuing legal name, legal obligor, brand at sale, distribution entity, jurisdiction, line of authority. It needs the group and control relationship, and must not replace the issuer with it. It needs a separate retention and reinsurance relationship, because “issuer” and “net risk bearer” answer different questions.

That structure makes a basic search possible.

Show every policy issued by this company, under this brand, through this MGA, in this province, whose trigger depends on this rating.

A brand hierarchy cannot answer that.

Neither can a group chart.

Legal form tells the system which fields exist

Stock, mutual, reciprocal: not decorative labels.

They change the control record we keep.

A stock insurer has share capital and an owner of record.

A mutual has no shares.

Policyholders’ surplus is its equity, and no ownership field exists. A reciprocal has subscribers and an attorney-in-fact, a separate manager paid a share of gross premium under the subscriber agreement.

The exchange is policyholder-owned; its manager is for-profit.

Not an edge case a parent-child hierarchy expresses well. Mutual holding companies add a stock holding company between members and the insurer.

Demutualisation makes the answer time-dependent.

The federal conversion regulations for mutual property and casualty companies came in 2015. Economical’s demutualisation closed in November 2021, immediately before Definity Financial, its new parent, listed on the TSX. A record storing “mutual” with no date preserves a stale conclusion as fact. Pooling is another reason to separate group and entity. An affiliate can cede to a lead company and assume back a pool percentage. Its statutory results are a slice of pooled business.

Not underwriting that entity conducted alone.

A rating can be assigned to the pool.

Attach it to every subsidiary and a familiar letter looks more precise than it is. In the US, Schedule Y and Form B describe group control and intercompany agreements, and pooling appears in the notes.

Canadian carrier data has its own divide.

OSFI’s register covers federal insurers and Canadian branches.

The AMF, FSRA, and BCFSA cover provincially incorporated insurers.

And every insurer, federal or provincial, still needs a licence in each province where it writes. OSFI supervises a federal company’s solvency; Ontario still licenses it to sell in Ontario.

One registry does not give us a carrier master.

Licence is a relationship, not a badge

“Admitted” is not a property of a company.

It is a property of the carrier and jurisdiction pair we are writing in, with separate authority per line. A carrier can be licensed somewhere and lack authority for our line.

Canada needs a different mental model.

We have no surplus-lines system.

Placement with an unlicensed insurer happens. It brings a 10% federal excise tax on net premiums, and provincial broker rules that differ by province.

Not an American eligibility-list regime.

Here, Lloyd’s Underwriters is federally authorised for non-life insurance and reinsurance, as an association rather than a company, with a chief agent and assets vested in trust in Canada.

“Non-admitted” in a Canadian workflow is not sophisticated shorthand.

It is an unexamined foreign field name.

A foreign insurer with a Canadian branch is another test.

The branch holds licences and appears in returns.

It is not a Canadian corporation with share capital. Its local solvency attribute is assets vested in trust here, and OSFI tests them against branch liabilities: the BAAT for property and casualty, the LIMAT for life.

Model the branch as a subsidiary and we will eventually produce a false balance-sheet comparison.

The capital fields we store likewise keep their basis.

Federal reporting has used IFRS 17 for years beginning on or after 1 January 2023: liabilities for remaining coverage and incurred claims, a risk adjustment, a contractual service margin.

Property and casualty capital runs on the Minimum Capital Test, whose 2026 guideline took effect in January; life runs on LICAT.

Not labels to hang on an enterprise brand.

They belong to the entity and perimeter they measure.

Guarantee-fund membership belongs there too.

PACICC is property and casualty, with exceptions such as some Ontario and Quebec mutuals covered by other plans. Assuris is life and health.

Neither follows the name we advertise.

The policy or complaint needs the Canadian member insurer, because a brand can span several companies and jurisdictions.

A rating needs an object

We load ratings into master data as a vendor name, a letter, and a date.

A useful start, the way a street address starts a reinsurance treaty.

A financial-strength rating is an opinion about a subject.

A.M. Best builds from balance-sheet strength anchored in BCAR, then operating performance, business profile, and risk management, before lift or drag from a parent.

BCAR is tested at several confidence levels.

The company assessed, the parent considered, and the pool reported are not interchangeable rows.

Their committee decides the outcome, not the analyst.

The entity is notified before publication. It can appeal only with material new information or a substantial clarification, and the window is counted in days: two calendar days for an existing rating, ten for an initial one.

It can request withdrawal of an interactive rating.

That leaves a withdrawal notice, and the last opinion in view.

Those states have to be dated and retained.

A blank field in our record is not the same as no rating.

Canada adds a particular regulatory boundary.

Designated rating organization status under National Instrument 25-101 sits with the Canadian Securities Administrators.

It governs procedure: a code of conduct, a compliance officer outside ratings, standards for quality, independence, transparency.

Five agencies hold it: DBRS, Fitch, Kroll Bond Rating Agency, Moody’s, and S&P. AM Best, the agency most insurance contracts name, is not among them.

Designation does not make a group letter apply to every insurer sharing a logo.

The record links agency, rating type, symbol, outlook, effective date, rated entity or pool, and source.

A financial size category belongs beside the rating where a counterparty relies on both.

A group relationship is context.

It is not a join condition.

Downgrades travel through clauses

Here entity resolution stops sounding like data hygiene.

Adverse reserve development can move an insurer from A- to B++.

The agency has published its opinion.

The damage reaches us through clauses written years earlier.

Reinsurance and fronting agreements can permit termination, commutation, or collateral below a threshold.

A- is the common line.

A cedant can lose regulatory credit unless collateral is posted.

An MGA’s agreement can require its capacity provider to hold A-.

A surety obligee may want A- plus a financial size category; US federal work wants Treasury listing.

Lenders can force-place when coverage falls below their required rating.

Broker security committees drop a market without waiting for anybody.

The scale looks gradual.

The way we use it is a switch.

A negative outlook is a directional view; AM Best generally means 36 months.

Under Review with Negative Implications is event-driven, signalling a potential change typically within six months.

Combine them in one “rating status” field and we learn nothing when a counterparty asks which happened.

The response isn’t a dashboard colouring a brand red.

It is a contract graph we key to the entity.

On a rating action, we find that entity’s policies.

The programmes it supplies. The treaties where it is cedant or reinsurer. The lender and surety requirements naming its rating. The broker lists. The licence assertions on open cases.

Then separate legal issuer from economic retainer.

A front’s downgrade and its reinsurer’s aren’t the same event.

Even where the policyholder never saw either name.

Make the reassuring sentence testable first

The board hears a group name, a rating letter, and a reassuring sentence about capital.

Our work is making it testable before it reaches the room.

Ask whether a downgrade, transfer, or complaint traces to the actual risk-bearing entity, without decoding a group chart by hand.

Ask whether that changes when the policy is issued by a branch, written through an MGA, pooled, or fronted to a captive.

Ask whether the system shows the source register, the jurisdiction, the dated rating action, and every clause borrowing it.

An unfashionable place to spend effort.

It doesn’t look like a new distribution platform.

It begins with interviews, old declaration templates, treaty schedules, and whoever knows why two near-identical company codes were never merged.

Good.

That is where our promise is hiding.

AI identifies entity names in declarations, compares group charts, links a withdrawal notice to a treaty schedule.

It should not decide which company is legally bound, whether a rating applies to a pool, or whether a trigger has fired.

Fluent entity matching is not contractual interpretation.

A system that can’t name the risk bearer can’t explain its security.