Life insurance: two statements, and why profit arrives decades late

A life insurer writes a flood of new policies, grows new-business premium 30%, and creates a vast amount of value for its owners - and its reported profit after tax falls. Not despite the growth; because of it. The faster it sells, the worse its bottom line looks in the year it sells, and a reader watching that bottom line would conclude the business is struggling at the exact moment it is winning. illustrative

This is not a quirk to wave away. When an insurer sells a twenty-year policy it incurs the cost of the sale now - commission, underwriting, the reserves it must set up - but earns the profit slowly, over the two decades the policyholder keeps paying. So the reported profit in any year is a strange, lagging thing: the slow emergence of profits from policies sold long ago, minus the up-front strain of policies sold this year. Grow fast and the strain dominates. The profit-and-loss account - the number every other business is judged on - is close to the least useful figure a life insurer files.

Which forces a different set of numbers. The value created is captured the day a policy is sold, in the value of new business. The worth of the whole book is captured in embedded value. And whether that value is real or illusory is captured in persistency - whether policyholders actually keep paying. This module reads a business where profit arrives decades late, and where the entire skill is seeing the value the reported profit hides.

Module 6 taught that profit is an opinion. Life insurance is the extreme case: reported profit is not merely an opinion but a genuinely poor measure of the year's economics, and a reader who brings a manufacturer's or even a bank's instincts will misread it completely - rewarding low growth, punishing high growth, mistaking the up-front cost of value creation for a loss.

Start with the structure, because reading the wrong statement is the first error. A life insurer files two. The Revenue Account belongs to the policyholders and holds the vast pool of premiums, investment income on that pool, benefits and claims, and the enormous change in policy reserves. This is where the business physically happens, and it dwarfs the shareholders' account. The Profit and Loss Account belongs to the shareholders, and its main input is a small surplus transferred out of the policyholders' fund each year, actuarially determined - set by an actuary, the specialist who prices insurance risk and its reserves. Reported profit after tax sits here, and it is small, lagging, and shaped by how much surplus the actuary released, not by how good the year was.

Three ideas replace it. Value of new business, or VNB, is the present value of the future profits expected from this year's policies - it captures up front the value the P&L will only reveal over twenty years. Embedded value is net worth plus the present value of the profits locked into the entire existing book - the best single measure of what the insurer is worth today. And persistency is the share of policyholders still paying after 13, 25, 37, 49 and 61 months - the honesty check on the other two, because a policy that lapses never delivers the value VNB assumed. Read the shareholders' P&L alone and you reach a confident conclusion about a business you have not actually looked at.

The mechanism is new-business strain. When a policy is sold, the insurer pays commission and underwriting and must set up reserves against the promise it has made - all now - while the profit emerges only as the years pass and the premiums keep coming. Each new policy is therefore a cost this year and a stream of profit for twenty. In a fast-growing year the strain from this year's policies weighs on reported profit while the emergence from old policies has not yet caught up, which is exactly why growth suppresses the bottom line.

Rather than wait two decades to learn whether a policy was worthwhile, actuaries compute its value up front. VNB is the present value of all future profit expected from this year's sales, net of that strain. Sum the same calculation over the entire in-force book, add net worth, and you have embedded value - the worth of the business today. Growth in embedded value - driven mainly by VNB plus the unwinding of prior years' value, last year's expected profit moving one year closer and so counted now - is the closest thing a life insurer has to a real annual profit, and it runs several times the reported figure. The reported PAT is not what the owner earned this year; the growth in embedded value is much nearer to it, which is why the accounting and the economics can point in opposite directions.

All of it rests on one assumption: that policyholders keep paying. Persistency measures whether they do - the percentage still active at 13 months, and further out at 61. High persistency means the VNB was earned honestly, on policies people wanted; low persistency means policies were sold that quickly lapse, so the VNB booked up front will never be realised. It is the single number that tells you whether the reported value is genuine or a projection that will quietly evaporate.

That reported profit measures the year's success is true for most businesses and structurally false for a life insurer. Set it beside its neighbours to see how far the sign reverses.

Hold the inversion at its sharpest: for a life insurer, reported profit and business success can point in opposite directions. A manufacturer that grows fast and profitably shows a rising profit; a life insurer that grows fast and profitably shows a falling one, because the value it created is deferred while the cost of creating it is immediate. This is a sign reversal, not a shade of emphasis. An investor who ranks life insurers by their price-to-earnings multiple - share price divided by annual profit - as they would a manufacturer is using a denominator that actively misleads, and will systematically prefer the slow-growing insurer to the fast-growing one. The right multiples are price-to-embedded-value and price-to-VNB, because those measure what the business actually did.

Take the composite life insurer's fifth year. The shareholders' P&L reports a profit after tax of ₹1,930 crore. Alone, on a business with an embedded value of ₹47,000 crore, that looks like a thin return - barely 4% on embedded value, worse than a fixed deposit. Value this insurer on reported earnings and you call it a poor business. illustrative

Now read what it did. The value of new business written this year was ₹3,980 crore - more than double the reported profit - at a VNB margin of 26%, up from 22% five years ago, so each rupee of new premium creates more value than before. Embedded value grew from ₹39,500 crore to ₹47,000 crore, a gain of ₹7,500 crore - nearly four times the reported profit - at an operating return on embedded value of 19%. That 19% is the real return the business earned on its worth, and it is excellent. The reported ₹1,930 crore was never the point; it was the small surplus the actuary chose to transfer to shareholders while the bulk of the value stayed locked in the policyholders' fund, emerging over decades.

Then check the honesty of it through persistency. Thirteen-month persistency is 88% and rising; 61-month persistency is 58% - most policyholders keep paying through the crucial first year, and well over half are still paying after five. That is a genuinely sold book, not a churned one, so the VNB booked will largely be realised. The mix is tilting toward protection (21% and rising) - pure life-cover policies with no savings element, the highest-margin and stickiest business. Every real metric says strong and improving; the one number that says otherwise is reported profit.

The habit to build: for a life insurer, ignore reported profit and the P/E almost entirely. Read VNB and its margin for the value written this year and whether it is improving; read embedded value and its operating return for the worth of the book and the return on it; read persistency across durations to confirm the value is real and not lapsing away. Value the insurer on embedded value and VNB, never on earnings.

Embedded value and VNB are actuarial calculations, only as good as the assumptions inside them. They project decades of persistency, mortality, expenses and investment returns, and management chooses those assumptions within a defensible range. An insurer that assumes slightly better persistency or slightly lower future expenses reports a higher VNB and embedded value with no change in the actual business - the same judgement lever that inflates a manufacturer's profit, moved into the actuarial model. VNB is a genuine improvement on reported profit as a measure, but it is not an observed fact; it is a projection, and the sensitivity of that projection to its assumptions is disclosed for a reason.

Nor does embedded value tell you what happens to the assumptions themselves. A book valued on 88% persistency is worth far less if a change in tax treatment, a mis-selling scandal, or a shift in the savings market pushes persistency down. Embedded value captures the worth of the in-force book under today's assumptions; it does not price the risk that those assumptions deteriorate. The value is real but conditional, and the conditions - regulation, competition, distribution economics - sit outside the number and must be judged separately.

And the framework, for all its sophistication, cannot make reported profit meaningful - it can only route around it, at a cost. The numbers that matter are actuarial, disclosed less often, and harder for an outsider to verify than a manufacturer's cash flow. You are trusting the actuary and the auditor more than in almost any other sector, because the value is inherently a projection of the future rather than a record of the past. Persistency and the sensitivity disclosures are the main external checks; beyond them the reader has less independent ground than usual - worth holding consciously rather than forgetting behind the elegance of embedded value.

With a life insurer, a sharp analyst pushes on the assumptions behind the value, because that is where reported strength is either earned or manufactured. The question sounds like this: "VNB margin rose 400 basis points over three years. How much of that is genuine mix shift toward protection, and how much is changes in your persistency and expense assumptions? And what's the embedded-value sensitivity to a 10% fall in persistency?" The analyst is separating a real improvement from a re-assumed one.

A good answer, verbatim-style.

"Of the 400 basis points, about 300 is genuine mix - protection went from 14% to 21% of new business, and it carries a far higher margin. Roughly 100 is a persistency assumption update, which we made because actual 13-month persistency has run at 87-88% for three years against the 85% we'd assumed, so it's catching up to reality, not getting ahead of it. On sensitivity, a 10% relative fall in persistency would reduce embedded value by about 4%, and we disclose the full sensitivity table on page 240. So most of the margin gain is mix, and the assumption change is backward-looking, not aggressive."

It splits the improvement into real mix and re-assumption, justifies the assumption change against actual experience, and gives the sensitivity with a source. It lets you judge how much of the reported value is earned.

An evasive answer, verbatim-style.

"We're delighted with our VNB margin expansion, which reflects our disciplined focus on high-quality, high-margin business and the strength of our multi-channel distribution. Our embedded value continues to compound at industry-leading rates, and our actuarial assumptions are prudent and reviewed regularly by our appointed actuary and the board. We're confident in the sustainability of our value creation."

Confident and unfalsifiable. It never splits the margin gain into mix versus assumption changes, never gives a persistency figure, offers no embedded-value sensitivity. "Prudent assumptions, reviewed regularly" is exactly what an insurer inflating VNB through optimistic assumptions would also say, and "industry-leading compounding" is a boast, not the decomposition the question asked for.

The follow-up nobody asks. "If you held all actuarial assumptions flat at last year's, what would VNB margin and embedded value be this year?" That strips out the re-assumption and shows the underlying business alone. If "prudent, reviewed regularly" is allowed to stand, an investor capitalises an assumption change as if it were operating performance. The silence is the tell - either the assumption changes are doing more of the work than management wants to admit, or the experience behind them would not survive the question.

The first trap is judging a life insurer on reported profit or its P/E, exactly as one would a normal company. Because profit emerges over decades and fast growth brings up-front strain, reported profit is a lagging, near-useless number, and ranking insurers by their P/E systematically rewards the slow-growing one and punishes the fast-growing one - the reverse of the truth. The insurer creating the most value can show the lowest reported profit. Reaching for earnings out of habit is the single most common error in the sector.

The second trap is reading premium growth as value growth. A life insurer can grow premium rapidly by pushing low-margin, single-premium market-linked products that lapse quickly, booking impressive top-line numbers while creating little embedded value and leaving a book that churns. Growth in premium is not growth in VNB, and VNB is not realised unless persistency holds. An insurer growing premium with a thin VNB margin and falling persistency is manufacturing the appearance of a good year; the value it reports up front will evaporate as the policies lapse.

The third trap is treating embedded value as a hard number rather than a projection. It is a far better measure than reported profit, but it is built on actuarial assumptions about persistency, expenses, mortality and investment returns, all chosen by management within a range. A rising embedded value driven by genuine mix shift and honest experience is real; a rising embedded value driven by quietly optimistic assumption changes is the same profit-flattering lever seen everywhere else, moved into a model most readers cannot audit. The check is to ask how much of the growth came from the business and how much from re-assumption, and to read the persistency experience and the sensitivity tables - because an embedded value taken on faith is exactly where a life insurer's story can be dressed up.