Insight
We look at what Ind AS 117 and India's coming risk-based capital regime mean for insurers, and what it actually takes to explain the number on the page, not just produce it.
Insights / Building the Foundations for Model Risk
India's financial institutions now run on models, whether they've noticed or not.
Ind AS 109 already puts credit risk, assumptions and forward-looking judgment into the numbers, for NBFCs and insurers alike. Ind AS 117 adds a second layer for insurers specifically, actuarial and accounting models built into every reported figure. And IRDAI's risk-based capital framework, still being finalised, is set to make capital itself a function of the insurer's own risk models.
The problem isn't whether these models work.
It's whether anyone can explain why they work, where they could be wrong, and what that means for the number on the page.
That is where model risk management becomes practical.
For many insurers, the difficult conversation starts when someone asks a simple question:
"Why did the model produce this number?"
The answer can quickly become complicated.
For Ind AS 109, it may involve probability of default, loss given default, exposure at default, staging, economic scenarios, forward-looking assumptions, overlays and data.
For Ind AS 117, it may involve fulfilment cash flows, discount rates, risk adjustment, Contractual Service Margin, grouping, cash-flow assumptions and actuarial methodology.
The model may be technically sound.
The problem is that technical soundness does not automatically make the model explainable.
A board member may want to understand the driver of a material movement.
An auditor may want to understand an assumption or methodology change.
Finance may need to explain why a reported number changed.
Risk may need to understand whether an output is behaving as expected.
The actuarial team may understand all of this.
But can the organisation explain it consistently, independently and with evidence?
That is a model risk question.
And it becomes harder as the number of models grows.
An insurer can have model validation reports, actuarial reviews, documentation and controls, yet still struggle to answer basic questions such as:
The issue is therefore not a lack of modelling expertise.
It is the ability to create a defensible line of sight from model to number to decision.
We start with a model inventory: every model that feeds into a reported number, across actuarial, finance, investment, risk and capital, with an owner and a record of when it was last checked. That sounds basic, but in our experience across financial services, most insurers can't answer basic questions about their own models on the spot: which version produced this quarter's number, what data actually went into it, whether a vendor quietly changed something upstream, a model version, an API, a data feed, without anyone noticing, or whether the model touched data it was never supposed to see in the first place. Not because the model is wrong, but because nobody wrote any of that down anywhere a board member or an auditor could find it.
From there we bring in someone from outside the team that built the model to test the assumptions rather than confirm them, and we push for outputs that can be explained in plain language: what's driving this number, what would move it, and what it doesn't account for. Every change to the model or its inputs gets logged and traced back to the reported figures it touched, so a methodology update doesn't quietly turn into an unexplained swing in profit six months later. Management and the board get one view of what's changed, what's still open, and what could go wrong, instead of five different answers depending on who they happen to ask.
The result isn't more paperwork. It's an organisation that can say, with evidence, exactly which model, which version, and which assumption produced the number on the page, and who's accountable if it turns out to be wrong. That's useful whether the model is behind Ind AS 109 today, Ind AS 117 reporting now, or risk-based capital tomorrow. It's the same foundation either way, whichever regulation brought you to the table.
Today's fixed solvency multiplier, the same number for every insurer regardless of risk. Risk-based capital retires it and replaces it with whatever the insurer's own models say.
How many times insurers have already modelled their own capital under the coming regime. This isn't a future scenario, it's already being rehearsed with real numbers.
Already the reason credit-risk models and forward-looking assumptions sit inside NBFC and insurer numbers today, before Ind AS 117 even arrives.
The date Ind AS 117 is scheduled to start counting for insurers, pending Gazette publication, adding actuarial models on top of what Ind AS 109 already put in place.
How long insurers must run old and new accounting side by side, unless IRDAI decides otherwise. Twice the models to reconcile and explain, not half.
The longest forbearance regulators are offering insurers who aren't ready, and only with a Board-approved plan showing they're getting there.
Ind AS 109 and Ind AS 117 are different requirements.
They are also different kinds of models.
But both create the same organisational challenge: important numbers increasingly depend on assumptions and models that need to be understood outside the people who built them.
That is the point at which model validation becomes model risk management.
Not because another regulation says so.
Because the business needs to be able to explain its numbers.
Talk to us about model risk management.
Get in touch →