What the money is for
An insurer selling motor cover collects premiums from a large group of people, most of whom will claim nothing, in order to pay the minority who will. The premium any individual pays is an estimate of their contribution to the group’s expected cost, plus the cost of running the operation, plus the cost of holding capital against the possibility that the estimate is wrong, plus a margin.
That is the whole structure, and several things people find strange about insurance fall out of it immediately. A premium is not a deposit and it is not a fee for a service delivered. A year with no claim is not a year in which you were overcharged; it is the ordinary case, and it is what makes the unusual case affordable.
Frequency and severity are modelled apart
Insurers do not really ask how much a customer will cost. They ask two questions and multiply them: how often is a claim likely, and how expensive is a claim likely to be. The two behave differently and respond to different variables, which is why a factor can push a price in opposite directions depending on which side it lands on.
Annual mileage mainly drives frequency, since exposure to the road is exposure to collisions. Vehicle value and repair complexity mainly drive severity, since they determine what a given impact costs to put right. A high-mileage driver in an inexpensive car and a low-mileage driver in an expensive one can arrive at similar premiums by entirely different routes.
Rating factors are correlations, and that is uncomfortable
A rating factor earns its place by predicting claims cost across a population, not by explaining it. Where a car is kept overnight predicts theft and damage because it captures something about traffic density, parking and local claim patterns. Occupation predicts claims for reasons that are partly about mileage, partly about hours driven and partly about things nobody can articulate cleanly.
This means a customer can be rated on characteristics they share with people they have never met and whose driving they do not resemble. It is a real limitation of the method and it is worth stating plainly rather than defending. Regulators in various jurisdictions restrict which factors may be used and how, and those restrictions differ and change, so what is permitted where you live is a local question.
Adverse selection is why the questions exist
Suppose an insurer stopped asking anything and charged everybody the average. Drivers who knew themselves to be low risk would find that price poor value and go elsewhere; drivers who knew themselves to be high risk would find it a bargain and stay. The remaining pool would be worse than the average, the price would have to rise, and the cycle would repeat.
That process is adverse selection, and it is the reason an insurer cannot simply choose to price generously. Every question at quotation is an attempt to distinguish risks well enough to avoid it. The insurer is not only estimating your risk; it is protecting itself against the fact that you know more about your driving than it does.
Why quotations for one person vary so widely
Each insurer has its own book of business, its own claims experience, its own reinsurance arrangements and its own appetite for particular kinds of risk. A company that already carries a large number of similar drivers has data on them and may price confidently; one with few may price cautiously or decline. Neither is a judgement about the individual.
Appetite shifts too, sometimes sharply, as insurers rebalance what they hold. A driver whose price is unremarkable at one company and startling at another has not been assessed twice. They have been assessed by two different models, built on two different histories, at two different moments in each company’s planning.
What a premium cannot tell you
A low price is a statement about how one insurer models a risk, not a verification that the cover is adequate. Two prices are only comparable if the cover, the excesses, the conditions and the add-ons behind them are comparable, and quite often they are not.
It follows that the useful comparison is never between two numbers alone. It is between two documents that happen to have numbers on them. That is more work than most people want to do once a year, and it is the only version of the exercise that means anything.
