Usage-Based Insurance

Usage-based insurance is auto coverage whose price is set from measured driving behavior rather than from the demographic and territorial characteristics insurers traditionally use as stand-ins for it. The National Association of Insurance Commissioners, the standard-setting body for state insurance regulators, describes it as an innovation "designed to better align driving behaviors with premium rates for auto insurance," and says it "examines driving habits (miles driven, speed, time of day and other factors) to determine insurance costs."

The naming is worth separating out, because two words get used interchangeably and mean different things. Usage-based insurance is the product. Telematics is one way of measuring for it: NAIC explains that "mileage and behavior-based measures are tracked using odometer readings or in-vehicle telecommunication devices (telematics)," and that "the basic idea of telematics auto insurance is that a driver's behavior is monitored directly while the vehicle is in use." The word "or" in that sentence is the reason the two are not synonyms: a program that prices on verified annual mileage read off the odometer is usage-based insurance with no telematics in it at all.

What conventional rating is doing, and why an observation beats a proxy. NAIC states that the single greatest influence on the rating process is claim frequency, and defines it as how often an insured event occurs within a group relative to the number of policies in that group, with the consequence that "persons sharing characteristics with high claims groups will be charged more for insurance coverage." That is group inference, and it is accurate on average and wrong about plenty of individuals. A careful driver in a territory with a poor claims record pays for the territory. Usage-based rating tries to replace that inference with a measurement of the individual, which is why it tends to benefit exactly the drivers whom the group model describes worst: low-mileage drivers, retirees, people who commute outside peak hours, and households whose second car barely moves.

What a device actually records. NAIC lists the elements telematics devices measure as being of interest to underwriters: miles driven; time of day; where the vehicle is driven, by GPS; rapid acceleration; hard braking; hard cornering; and air bag deployment. It adds that "the level of data collected generally reflects the technology employed and the policyholders' willingness to share personal data," and that the insurer then assesses the data and charges premiums accordingly. Two of those items are worth pausing on. Location by GPS is a continuous record of where a household goes, which is a materially different disclosure from a mileage count. Air bag deployment is a crash detection signal, which some programs use to trigger emergency response and which is also a record of an event the insurer would otherwise learn about only from a claim.

The questions worth asking before enrolling, none of which have a universal answer. Whether the program can only reduce the premium or can also increase it is a term of that program and of the rate filing the insurer made with the state, so it is a question for the insurer rather than something that can be answered generally. So is how long the monitoring period runs, whether the rating persists after it ends, what happens to the data afterwards, whether it is shared with anyone else, and whether it can be obtained in a claim dispute or by subpoena. A program priced purely on verified mileage collects far less of a household's information than a behavioral program does, and that is a real distinction rather than a technicality.

Where usage-based rating sits among the other things that set the price. It is one input, alongside the coverages and limits chosen, the deductible, the vehicle, the territory, the driving record and, in states that permit it, a credit-based insurance score. Nothing about it changes the policy: the same liability limits, the same physical damage coverage, the same exclusions. What changes is the number on the bill, which makes the decision a pricing and privacy question rather than a coverage question.

The policyholder enrolls, and the insurer begins collecting the agreed measure, which may be a periodic verified odometer reading, a plug-in device, a factory connection in the vehicle, or a phone application. After a monitoring period the insurer applies the result to the rate, either as a discount off an otherwise conventional premium or as the basis of a per-mile price. Coverage terms are unchanged throughout.

A hypothetical, to show what a per-mile structure does. Suppose a program charges a fixed base of $32 a month plus 6 cents a mile, and a conventional policy with identical coverage would cost $1,050 a year. A driver covering 5,000 miles a year pays twelve months of base, which is $384, plus 5,000 times $0.06, which is $300, for a total of $684. That is $366 less than the conventional policy. A driver covering 14,000 miles a year pays the same $384 of base plus 14,000 times $0.06, which is $840, for $1,224, which is $174 more than the conventional policy. The break-even is where the mileage charge equals the $666 difference between the conventional premium and the annual base, which is 11,100 miles. All three figures are invented for the arithmetic; real programs differ by insurer, state and coverage.

The shape of that result is the general lesson and it survives the invented numbers. A per-mile structure is a straightforward win below the break-even and a straightforward loss above it, and the break-even is computable in advance from the program's own terms and last year's odometer. A behavioral program is harder to evaluate in advance, because the score depends on driving the policyholder has not done yet and on a model the insurer does not publish.

Pros

• It prices the individual rather than the group, which helps the drivers whom group inference describes worst.
• Low-mileage households, retirees and lightly used second vehicles are the clearest beneficiaries, and the saving is estimable in advance from a per-mile program's published terms.
• Feedback on braking, acceleration and time of day gives a driver information they otherwise never see about their own habits.
• It does not alter coverage. Limits, deductibles and exclusions are the same policy they would otherwise be.

Cons

• The detailed programs collect a continuous record of where a household drives, and regulators note the data collected reflects how much the policyholder is willing to share.
• Whether a program can surcharge as well as discount is a term of that specific program, so it has to be asked rather than assumed.
• A behavioral score depends on a model the insurer does not publish, so the outcome cannot be estimated in advance the way a per-mile price can.
• Driving that is safe but scores badly, such as necessary hard braking in city traffic or a night shift, is penalized by a measure that cannot see context.
• High-mileage drivers are the natural losers of a mileage-based price, which is the arithmetic working as designed rather than a defect.