The net premium was never the whole bill
In the previous guide you priced a policy with the equivalence principle: set the premium so that, at issue, the expected present value of the premiums you collect exactly equals the expected present value of the benefits you will pay. The number that comes out is the net premium (also called the net level premium or the pure premium). It is a beautiful, clean idea — but notice what it pays for and what it quietly ignores.
The net premium pays for *exactly one thing*: the claims, on average, discounted for interest and survival. That is the pure cost of bearing the risk — nothing more. But a real insurer does not run on good intentions. Before a single dollar of claim is ever paid, money has already flowed out the door: the agent who sold the policy must be paid a commission, the underwriters and clerks who issued it draw salaries, the building has rent and electricity, the regulator wants its filing fees, and the government wants premium tax. None of this appears in the equivalence principle. A company that charged only the net premium would be insolvent within a year — not because its claim estimates were wrong, but because it forgot to charge for the cost of *being a company*.
The four costs a premium must cover
It helps to lay out, plainly, everything the price you pay must shoulder. There are four buckets. First, the claims themselves — this is the net premium, the only bucket the equivalence principle already filled. Second, commissions: the chunk paid to the agent or broker who brought the business in, usually a large slice of the first year's premium and a smaller trickle thereafter. Third, administration and overhead: salaries, systems, rent, the cost of mailing statements and answering the phone for thirty years. Fourth, taxes and a profit margin: premium taxes and levies that governments charge, plus the return the company's owners require for putting their capital at risk.
These costs do not all behave the same way, and that matters enormously for how we charge for them. Some are *one-off and front-loaded*: the agent's commission is paid almost entirely when the policy is sold, and the medical exam happens once. Others are *recurring*: the cost of administering the policy lands every year for as long as it stays on the books. And some are *proportional to the premium itself*: premium tax is a percentage of what you collect, so it scales automatically with the price. An honest gross premium has to recognise each of these rhythms separately rather than smear them into one fudge factor.
That heavy front-loading of commissions is not just an accounting curiosity — it is the seed of one of the deepest puzzles in this whole rung. Because the insurer pays so much cash out in year one but collects level premiums spread evenly across the years, the policy actually *loses money at issue*. We packaged that pain into the loss-at-issue random variable and the deferred acquisition cost, and it is the reason reserves exist. For now, just hold the thought: the lumpy timing of expenses is going to come back and shape everything.
Expense loadings: how the extras get charged
An expense loading is simply the amount we add to the net premium to pay for everything outside the claims. The art is matching the *shape* of the loading to the *shape* of the expense. Actuaries usually split expenses into three natural types. A per-policy expense is a flat dollar amount — say, the cost of issuing and mailing one policy, the same whether the sum insured is small or large. A percent-of-premium expense — commission, premium tax — scales with the premium you charge. A percent-of-benefit or per-unit expense scales with the size of the cover, like the cost of underwriting a larger sum at risk. Choosing the right type for each cost is what makes the loading fair across big and small policies.
Now the equivalence principle gets a small but profound upgrade. Instead of "premiums = benefits" we write "premiums = benefits + expenses", all in expected present value at issue. The premium that solves *this* fuller equation is the gross premium, also called the office premium because it is the price the office actually quotes. Notice the lovely self-consistency: percent-of-premium expenses depend on the very premium we are solving for, so the gross premium appears on both sides of the equation. We simply solve for it — a one-line piece of algebra, not a circular trap.
A tiny gross-premium build-up (annual premium, one unit of cover) Net premium (pure cost of claims) 100.00 <- from equivalence principle Expense loadings, by type: Per-policy admin (each year) + 8.00 Percent-of-premium: commission 10% + tax 3% + 13% of gross premium G Set premiums = benefits + expenses, in PV at issue: G = 100 + 8 + 0.13 * G G - 0.13 G = 108 0.87 G = 108 G = 108 / 0.87 = 124.14 <- the gross (office) premium Loading = 124.14 - 100 = 24.14, i.e. about 19% above the net premium. (Note G sits on BOTH sides because tax & commission are % of premium.)
Where does the profit live?
We have covered claims, commissions, admin, and taxes — but a company that merely breaks even on every policy gives its shareholders no reason to put capital at risk. The missing piece is the profit margin, often called the underwriting profit provision or the profit loading. It is the deliberate sliver of the gross premium that, in expectation, is left over after every promise is kept and every cost is met. Without it, no one funds the insurer, and with no capital there is no insurer at all.
How big should that margin be? It is not pulled from the air — it is anchored to the *cost of capital*. Regulators force the insurer to hold a buffer of capital against bad years, and that capital could otherwise have earned a return elsewhere. The profit loading is essentially the rent on that locked-up capital, so a riskier product — one demanding a bigger buffer — earns a bigger margin. This is the honest version of the cynical-sounding truth that the price you pay exceeds the pure cost of your risk: the excess is not greed, it is the price of the safety the company must keep on hand so it can still pay when a bad year arrives.
There is a subtle but important honesty here. The margin is only an *expected* profit. On any single policy the insurer might still lose money — the policyholder might die early, or claims might cluster. The profit loading does not guarantee a gain on every contract; it tilts the odds so that across a whole portfolio, after the dust settles, there is on average a return for the risk taken. Confusing expected profit with guaranteed profit is exactly the kind of error this field trains you to never make.
A risk-based alternative: the portfolio percentile premium
Loading the net premium with explicit expenses and a margin is the classical, build-it-up-by-parts approach. But there is a second, more statistical way to ask "how much extra should we charge?" — and it speaks directly to risk rather than to a fixed margin. It is the portfolio percentile premium. The idea: forget about naming a profit percentage, and instead pick a premium high enough that the *entire portfolio of policies* will, with some chosen high probability — say 95% — collect enough to cover all its benefits and expenses.
This is where two earlier rungs quietly pay off. The total loss on a portfolio is a sum of many individual policy losses, and by the central limit theorem that total is approximately bell-shaped, with a mean and a standard deviation you can compute. The percentile premium just sets the price so that the expected total loss plus enough standard deviations to reach your chosen percentile comes out to zero or below. Charge for the average loss, then add a cushion sized in *standard deviations* — the more certain you want to be, the more cushions you stack, and the higher the premium climbs.
Two honest caveats keep this powerful idea grounded. First, it is a *portfolio* construction: the cushion per policy shrinks as the portfolio grows, because risks diversify — the same statistical magic behind risk pooling, now showing up in the price. A percentile premium for a portfolio of ten policies is far heftier per policy than for a portfolio of a million. Second, the method leans on the bell-shape being a good description, and it watches a percentile — say the 95th — not the very worst case. It says nothing about how catastrophic the unlucky 5% tail could be. For ordinary life portfolios that is usually fine; for heavy-tailed, catastrophe-exposed risks it can lull you into a false calm. A percentile is a useful line in the sand, not a guarantee against the deep tail.
Why the price always exceeds the pure cost of risk
Step back and the whole picture clicks into place. The net premium answers a pristine mathematical question — what does the risk cost on average? The gross premium answers the messier, truer question — what must a real company charge to keep that promise, pay everyone who makes it possible, satisfy the tax collector, and still earn a fair return on the capital it risks? The first number is the foundation; the second is the foundation plus the building that sits on it.
So when a friend grumbles that insurance is a rip-off because the company "keeps more than it pays out", you now hold the honest reply. Yes, the gross premium sits above the expected claims — it must. Part of the gap pays the people and systems that turn a promise into a cheque that actually arrives; part is the rent on the capital standing guard against the bad year. The pure cost of risk was never the whole cost of *transferring* that risk to someone strong enough to carry it. That difference is not a trick — it is the price of the safety itself.