The law of large numbers is the mathematical principle that makes insurance possible. It states that as an insurer adds more policyholders to its pool, the average actual losses will converge toward the expected (predicted) losses. With 100 policies, actual claims might swing wildly from year to year. With 100,000 similar policies, the average cost per policy becomes highly predictable, and that predictability is what allows insurers to set stable premiums.
How the Math Works
The core mechanism is surprisingly simple. The variability an insurer faces, measured by the standard deviation of average claims, decreases by the square root of the number of people insured. If one person’s annual claims have a standard deviation of $5,000, insuring 10,000 similar people brings the standard deviation of the average claim down to $50. Insure a million people and it drops to $5. The individual risk hasn’t changed at all, but the group’s average becomes extremely stable.
This is why a single person can’t self-insure against a house fire. Your house either burns or it doesn’t, and the outcome is binary and devastating. But an insurer covering 500,000 homes can predict with narrow precision how many will burn each year, even though it can’t predict which ones. The gap between what the insurer expects to pay and what it actually pays shrinks as the pool grows.
Two Requirements the Pool Must Meet
The law of large numbers doesn’t work automatically just because an insurer writes more policies. Two conditions must hold.
First, the risks must be independent. One policyholder filing a claim shouldn’t make another policyholder more likely to file one. A car accident in Phoenix has nothing to do with a car accident in Boston, so auto insurance pools satisfy this requirement well. Flood insurance in a single river valley does not, because one flood damages hundreds of homes simultaneously.
Second, the risks must be similarly distributed, meaning the policyholders in the pool share comparable risk characteristics. You can’t lump 20-year-old drivers and 80-year-old drivers into one undifferentiated pool and expect the average to stabilize in a useful way. Insurers handle this by segmenting policyholders into groups with similar profiles, then applying the law of large numbers within each group.
From Prediction to Premium
Once an insurer can predict average losses reliably, it can price policies. The foundational idea in actuarial science is called the principle of equivalence: premiums should be set so that the money coming in from policyholders and the money going out in claims are balanced on average. As the number of independent policyholders grows, the mean balance per policy converges toward zero, meaning the insurer collects just enough to cover what it pays out, plus its operating costs and profit margin.
In practice, an insurer looks at historical claims data for a large group of similar risks, calculates the expected loss per policy, then adds a loading for administrative expenses, a reserve margin for years when claims run higher than expected, and profit. The larger and more homogeneous the pool, the smaller that reserve margin needs to be, because the insurer faces less uncertainty. This is one reason large insurers with millions of policies can sometimes offer lower premiums than smaller competitors covering the same risks.
When Large Numbers Aren’t Enough
Simply growing a portfolio doesn’t guarantee stability if the underlying risks aren’t well matched. One case study illustrates this clearly: an insurer that acquired blocks of policies from other companies saw its book grow substantially, yet loss ratios stayed volatile, swinging between 45% and 180% annually on an initial portfolio of 10,000 policies. The problem was that each acquired block had different underwriting standards, coverage forms, and risk profiles. The pool got bigger, but it never became homogeneous, so the statistical benefit never materialized.
The lesson is that the law of large numbers cannot fix bad underwriting. If an insurer is attracting a disproportionate share of high-risk customers (a problem called adverse selection), writing more policies just multiplies the problem. Statistical principles require consistent risk evaluation across the portfolio.
The Problem of Correlated Catastrophes
The independence requirement breaks down most dramatically with catastrophic events. A hurricane doesn’t damage one house at random; it damages thousands of houses in the same region on the same day. A pandemic doesn’t send one person to the hospital; it overwhelms entire healthcare systems simultaneously. These are correlated risks, and adding more policyholders in the same region or the same sector doesn’t reduce variability. It increases total exposure.
Climate change is intensifying this challenge. Risks like wildfire, flooding, and extreme heat cascade across systems in nonlinear ways, creating what researchers describe as intersystemic systemic risks. The cause-and-effect relationships involve poorly understood tipping points, and the distribution of negative impacts is hard to predict. Traditional insurance modeling, built on the assumption that last decade’s loss patterns will roughly repeat, struggles with risks that are accelerating and interconnected. This is a key reason some insurers have pulled out of high-risk markets like coastal Florida or fire-prone areas of California.
Rare Events in Large Pools
There’s an ironic flip side to the law of large numbers. When a pool gets truly enormous, events that are extremely rare for any individual become nearly certain to happen to someone. An insurer covering 10 million homes will see freak losses every year: the house struck by a meteorite, the fire caused by a bird dropping a lit cigarette through a window. Actuaries call this the law of truly large numbers, and it means that large portfolios must budget for bizarre, low-probability claims as a routine cost of doing business. This matters especially for specialized pools covering low-frequency but high-severity risks, such as nuclear energy or aerospace, where a single event can generate enormous losses.
Why It Matters for You
Understanding this principle explains several things you encounter as an insurance buyer. It’s why insurers ask so many questions about your age, health, driving record, or property details: they’re trying to place you in a pool of similar risks where the math works. It’s why niche insurance products (covering unusual hobbies, rare collectibles, or unconventional businesses) tend to cost more relative to the coverage amount, because the pool is smaller and predictions are less reliable. And it’s why your rates can spike after a widespread disaster even if you personally didn’t file a claim. When a correlated event hits many policyholders at once, it reveals that the pool wasn’t as diversified as the pricing assumed, and premiums adjust to reflect that reality.