Credit Data Cannot Decide Whether a Loss Is Political

A country score is useful on a Monday morning. It tells the credit team to look harder at a stressed jurisdiction. It tells our treasury the local currency is getting difficult. It makes our trade desk uneasy for sensible reasons. It cannot decide what caused a loss on Thursday afternoon.

A country-risk score is an early-warning tool, not a coverage determination wearing a dashboard.

Trade-credit insurance and political-risk insurance meet at an overseas receivable, then immediately ask different questions. One asks whether a named buyer failed to pay under the policy’s commercial machinery. The other asks whether a named political peril caused the loss. State action. Transfer restriction. Violence. That seems obvious until a platform holds one buyer file, one country alert, one queue, one red flag. Then an unpaid invoice starts acquiring a political explanation, because the country is under strain. We have made a correlation feel like an instruction. That is the mistake to prevent.

The same failed payment can begin in two systems

Trade credit begins with a buyer relationship we underwrote. A whole-turnover policy makes the insured declare all eligible sales, not just the accounts it distrusts. A single-buyer policy covers one relationship and skips that machinery. The distinction is commercial, and it changes the record available later. Declared turnover, buyer identity, territory, limit, delivery history: not administrative residue. Conditions of the cover we wrote. The territory is a schedule, and the governing test is normally the buyer’s domicile. Not the delivery address. A shipment to a Canadian warehouse invoiced to an unlisted parent may be uninsured. Poor moment for our customer master to decide the warehouse address is close enough. The buyer limit isn’t an invoice flag either. It is a revolving cap on what we have outstanding. A discretionary credit limit sits under a hundred thousand, and needs evidence held before shipment. A trading record, an agency report, a bank reference, as the wording demands. Above it, a named limit is our underwriting decision. So we keep the limit’s effective date, the amount, the authority route, the evidence.

When we reduce a limit prospectively, deliveries already made under the old limit stay covered.

Later deliveries face the reduced figure. A monitor can surface the change in seconds.

It can’t make a later delivery an earlier one. Political risk begins somewhere else entirely. It is named-peril cover.

Never all-risk.

The closed list holds confiscation, expropriation, nationalisation, deprivation.

Then inconvertibility and transfer restriction, political violence, sometimes breach of contract by a state entity.

Commercial default sits outside it, alarming country or not.

Our systems touch at the receivable.

They should not merge their causes there.

A payment problem is not yet a political event

The useful first split is embarrassingly plain.

Did the buyer fail to pay, or was payment prevented by a covered state-caused event?

We need room for both, because the answer may be commercial, political, mixed, or not yet established.

For trade credit, insolvency is a defined legal event rather than a financial mood.

A bankruptcy or receivership under the Bankruptcy and Insolvency Act, a proposal, or a CCAA filing can trigger the insolvency route.

A buyer failing but filing nothing takes protracted default.

That route runs its own clock.

Protracted default runs five or six months from the due date before we can file.

The account may need notice at thirty days past due.

One form wants notice within thirty days of the trigger.

A maximum extension permits sixty days beyond terms.

A salesperson agreeing a longer plan changed the insurance result.

Nobody touched the claims screen.

Stop-shipment turns our policy into a credit instruction.

Past the notification trigger, further deliveries can be uninsured.

Some wordings say they prejudice the earlier debt.

A queue optimised for cash wants the conversation open.

The workflow has to know when that instinct lost its authority.

The computation we run stays commercial unless facts say otherwise.

It subtracts credit notes, allowances, contras, disputed invoices, balances above the limit at delivery, and goods shipped after the stop date.

A quality dispute suspends non-payment cover until judgment or agreement resolves it.

No country model decides whether the steel met spec.

Inconvertibility is not a bad exchange rate

Currency inconvertibility creates the most tempting false merge.

The buyer may hold local currency.

The exporter is unpaid in the currency it needs.

The country may be in a currency crisis.

Every dashboard we own agrees this is bad.

But inconvertibility insures inability to convert through legal channels.

Not the rate at which local money trades.

Devaluation and depreciation are excluded.

The proof is a formal conversion request to a central bank or authorised dealer.

Then a dated refusal or unanswered receipt, and documented attempts through the waiting period.

That evidence isn’t a data-quality nuisance.

It is the claim itself.

A parallel-market rate is interesting and legally irrelevant.

We need the application, the institution, the currency tendered, every response, the dates.

We should not let a country score quietly replace those records with an inference.

The political-risk workflow decides beneficiary and asset early.

For an investment, a policy measures equity against net book value, or a declining insured amount.

For lenders, outstanding principal plus accrued interest.

Export Development Canada writes political-risk insurance for Canadian investors abroad under the Export Development Act.

Its mandate and terms are part of eligibility.

Not a generic Canadian-country flag.

Government action has to be classified, not admired

Expropriation isn’t our label for every hostile act.

A confiscation claim needs an act attributable to the host government that deprives the investor of fundamental rights, or of use and benefit.

The deprivation continues through the waiting period, three months to a year.

A sequence of measures can be creeping expropriation.

A difficult quarter is not yet one.

The police-powers carve-out makes the classification more exacting.

Bona fide regulation of general application is excluded.

A measure aimed at every producer differs from one aimed at a foreign-owned concession.

An event feed calls both a regulatory change.

Only one of them supports a political-risk claim.

Political violence needs the same discipline.

War, insurrection, revolution, terrorism, sabotage: distinct definitions.

Sub-limits and exclusions can attach differently to each.

A loss caused by ordinary criminal violence does not become terrorism because a country score is high.

Nor because a news feed wrote a caption.

The temptation is to hand AI the ministerial order, the correspondence, and the reports, and let it tell operations what the loss is.

Let it extract, compare, and flag inconsistencies.

Let it build us a chronology.

It should not hold coverage authority.

“The model found this politically likely” won’t survive review.

One country signal, and two separate authorities

There is good reason to share country information.

A payments crisis moves a buyer’s outlook, hardens conversion requests, prompts a territory review, touches sanctions screening.

Duplicate entry is not a control.

The shared component holds the country event, source, date, currency, legal entity, and confidence in the evidence.

The authority splits after that.

Trade credit owns buyer monitoring, limits, terms, overdue notice, stop-shipment, collections, and the default path.

Political risk owns the named peril, state attribution, wording, waiting period, local remedies, consent to settle, and rights against the host state.

Sanctions sit across both and simplify neither.

Territory clauses can exclude payment that would breach Canadian, American, European, or UN measures.

Political-risk limitation clauses can suspend cover for the same reason.

Here, the Special Economic Measures Act and the Justice for Victims of Corrupt Foreign Officials Act may both apply.

A payment path can close while the coverage analysis holds.

For officially supported export credit at two years or more, the OECD Arrangement adds another constraint.

Country classifications, minimum premium rates, maximum repayment terms, a fifteen per cent minimum cash payment.

It doesn’t turn investment insurance into trade credit.

It doesn’t decide why a buyer defaulted.

The board conversation isn’t about our country-risk dashboard.

It is whether our record explains alert to authority.

Ask for the buyer event, the state event, the trigger, the notice and waiting-period clock, and whoever owns the next escalation.

That record also protects the lender.

A loss payee endorsement or assignment directs money to a lender, with PPSA registration or RDPRM registration in Quebec.

It does not cure the insured’s missed notice.

Non-vitiation wording can change the lender’s position.

A negotiated condition, not an effect of an integration.

Keep the queues apart until the facts earn a merger

The architecture I want isn’t one collections experience with a cause field.

It is two workflows sharing evidence, not authority.

A commercial-credit case and a political-cause case, linked by receivable and country event, each able to close alone.

That feels duplicative.

Cheaper than arguing whether an unreported overdue account was an inconvertibility claim.

Or whether a buyer’s quality defence quietly became political violence because the country turned red.

Sometimes both files stay live.

Not a systems failure.

It is honest uncertainty under a contract.

We keep the facts, run the right clocks, and stop one workflow asserting for the other.

A failed payment becomes an insurance question only once we can say why it failed.