Nine Questions on Insurance Risk and the Move to Full Repricing

*A conversation with Tatiana Zebaze, Principal Product Manager, Risk for Insurance.*

Insurance risk has changed more in the last decade than in the twenty years before it. We sat down with Tatiana Zebaze to talk about what’s driving that change, why regulators and rating agencies increasingly expect a fully recalculated portfolio rather than an estimate, and what it means for mid-market insurers today.

Q1. How have insurance risk regulatory frameworks evolved over the last ten years and what have you seen change?

The last decade has been a shift from formulas to actual portfolios. Ten years ago, a lot of insurance capital and reserving was factor-based, applying a prescribed percentage to a book value and you had your number. That world is largely gone. Solvency II set the tone in Europe with market-consistent valuation and a risk-based capital charge, and everyone else has moved in the same direction: the US toward principle-based reserving and expanded scenario testing, Asia toward risk-based capital regimes in Hong Kong, Singapore and beyond, and the rating agencies toward far more granular, scenario-driven capital models.

Three changes stand out:

First, the scenario sets keep growing and updating: the prescribed interest-rate paths that were seven a few years ago are now ten, and several regional regimes are refreshed every year or two.

Second, the expectation moved to instrument-level detail, including look-through into structured products and funds, rather than portfolio averages.

And third, transparency: regulators and rating agencies increasingly want to see the actual calculation, with every assumption traceable, not a summary you are asking them to trust. Taking together, the direction of travel is unmistakable: from estimate to recalculation.

Q2. What do you think are the biggest challenges in the market right now? Especially for mid-market insurers?

The mid-market is caught in a scale gap. The largest carriers responded to all of this by building in-house quant teams and heavy infrastructure. A mid-market insurer faces the same regulatory and rating-agency expectations, and increasingly the same portfolio complexity, without that scale of people or systems.

So, they are squeezed from two sides. Their portfolios have become more complex (private credit, structured products, derivatives, alternatives) at the same time the regulatory ask has become more demanding: more scenarios, more frequently, at instrument level, with an audit trail.

Meanwhile much of the tooling they rely on is a mix of spreadsheets and legacy analytics designed for a simpler book and a slower cadence. Data sits in silos (positions in one system, analytics in another, actuarial in a third) and reconciling those consumes the very teams that are hardest to staff. Quants are difficult to hire and retain, and a smaller insurer cannot stand up a large model-risk function.

The result is that approximation used to be good enough and increasingly isn’t, but the traditional way to close that gap was a budget and a headcount mid-market insurers simply don’t have. That is the real problem to solve.

Q3. Why is full repricing more important than approximation, and why now?

Approximation was a sensible shortcut for a simpler world. If your book is mostly vanilla bonds and the regulator accepts an estimate, then sensitivities (a duration and convexity adjustment, some bucketing) get you close enough. Two things broke that. The portfolios stopped being simple, and the regulators and rating agencies stopped accepting estimates.

The technical heart of it is this: an approximation is a shortcut around the actual calculation, and it fails precisely where the risk lives. Duration and convexity are a local estimate, and the scenarios regulators prescribe are large deliberately. For a bond with a call, a structured product with prepayment, or anything path-dependent, a bucketed estimate can be materially wrong exactly under the big moves you are being asked to test. So, the shortcut is least reliable and now it matters most.

Full repricing means you don’t scale a base number; you revalue every position from its own terms under each shocked market state and let the real non-linearity show up. Why now? Because the scenario sets got larger, the assets got less linear, and the appetite for “trust our estimate” fell. When you take a number to your board or your regulator today, you increasingly need to show it is the portfolio recalculated, not a proxy.

Q4. In plain terms, what does “repricing from scratch” require under the hood?

Three things must come together:

First, a faithful model of every instrument, not a label, but its actual mechanics: the coupons, the call schedule, the amortization, the prepayment behavior, the look-through into a fund or a structure.

Second, the market data to construct each scenario (the curves, spreads, volatility surfaces and fixings) and the machinery to shock them the way the scenario prescribes.

Third, a valuation engine that takes every position and re-runs it under each shocked market state, producing the cash flows and the values.

The hard parts are the ones people underestimate. Coverage is the first: you must model the awkward instruments, the private placements and structured credit and derivatives, not just the vanilla bonds, because that is where the exposure increasingly sits. Market data is the second: sourcing and shocking curves and volatilities correctly per scenario is real work. And then doing it at scale, quickly, repeatably, with a complete audit trail so every number traces back to its inputs and assumptions.

The piece I would add is where the data comes from. Ideally this runs on the same book of record the insurer already uses for its accounting; so the risk numbers reconcile to the numbers they report, with no separate data build. And because it is a genuine revaluation, one engine serves every framework: you change the scenario set, not the system. One data set, every filing, no re-keying.

Q5. Private assets are now a large share of insurance portfolios.  How does risk management keep up with that trend?

This is one of the biggest structural shifts in insurance balance sheets, and it is where approximation breaks down hardest. Insurers — especially in life and annuity — have moved heavily into private credit, direct lending, structured credit and fund investments in search of yield. The difficulty is that these assets don’t arrive with a market identifier and a clean daily price.

Each type needs something specific. A private placement has bespoke terms and no quoted mark, so you must capture its actual terms and conditions and value it from them; that is a data capability, not a market feed. Structured credit needs proper cash-flow modelling with collateral, prepayment and default assumptions, not a single number. Fund investments need look-through into the underlying holdings, because you can’t shock a black box; you must see what it is inside. And some private debt is drawn down over time on a commitment, which means the undrawn part is a liquidity obligation you need to plan for, not a footnote.

The wrong answer (and, unfortunately, a common one) is to proxy private assets crudely or leave them out of the stress entirely. That is precisely the part of the book where the risk now concentrates. Keeping up means treating these instruments with the same rigor as a listed bond: real terms, real look-through, real credit modelling.

Q6: The number you file versus the number you show the board, why does it matter that risk runs on the same data as the books?

Why: Gives the reconciliation / book of record point its own beat. It is arguably the strongest differentiator and currently only sits implicitly inside Q4.

Q7: Beyond meeting the rules, how are the most sophisticated insurers using full repricing?

Why: Lets the compliance to decision, resilience-testing story stand on its own rather than being compressed into the closing question.

Q8: What should a mid-market risk leader look for when evaluating a platform?

Why: A buyer’s-guide framing that lets you plant the evaluation criteria (instrument coverage, running on the book of record, one engine across frameworks, auditability, private-asset handling) as the reader’s own checklist.

Q9. What are you most excited about in the rest of the year?

What I am most excited about is the shift from compliance to decision-making. For years, running these calculations was a pass/fail exercise (did you clear the regulatory bar or not). Once you can genuinely reprice the whole portfolio, the same capability becomes a forward-looking tool. A risk team can take the prescribed scenarios and flex them (move the magnitude, the speed, the starting point) to ask how the book holds up if rates evolve differently than the regulation assumes. They can build their own cross-asset scenarios from scratch. Compliance becomes the floor, not the ceiling.

The second thing is access. The capability I have described used to require a big team and a big budget, which meant mid-market insurers were effectively priced out of enterprise-grade risk. Bringing that within reach (instrument-grade, defensible, across every framework, without having to build it in-house) is the part of the work that genuinely motivates me. There is a lot landing this year on both fronts, and I am looking forward to putting it in front of people.