Pay-As-You-Drive Insurance: Actuarial Considerations

Mileage-Based Rating

Pay-as-you-drive (PAYD) insurance charges premiums based on the number of miles driven, reflecting the strong and well-documented relationship between driving exposure and accident frequency. Actuarial studies consistently show that miles driven is one of the strongest predictors of claim frequency, with doubling annual mileage roughly doubling the expected number of at-fault accidents. This relationship is intuitive: more time on the road creates more opportunities for collisions, regardless of the driver's skill or caution.

Traditional annual premium structures charge low-mileage and high-mileage drivers the same rate within each rating class, or at best use self-reported mileage brackets with limited verification. This creates a cross-subsidy where low-mileage drivers effectively pay more per mile of exposure than high-mileage drivers. A driver covering 5,000 miles per year pays the same annual premium as a driver covering 15,000 miles per year (assuming identical values for other rating variables), despite presenting roughly one-third the exposure. PAYD programs eliminate this cross-subsidy by using odometer readings, telematics devices, or connected car data to verify actual mileage and charge accordingly.

The Mileage-Frequency Relationship

The statistical relationship between mileage and claim frequency is well established but not perfectly linear. Research using telematics data has shown that the marginal risk per additional mile decreases slightly at higher mileage levels. Several hypotheses explain this non-linearity. High-mileage drivers accumulate more driving experience, which may improve their skills. A larger proportion of high-mileage driving may occur on highways, which have lower accident rates per mile than urban streets. Additionally, driver fatigue effects may plateau rather than increase proportionally with miles driven.

Actuaries model this relationship using various functional forms. A simple linear model assumes that expected claim frequency equals a base rate plus a slope coefficient times annual miles. A power model assumes that frequency is proportional to miles raised to an exponent less than one (capturing the diminishing marginal risk). A log-linear model includes log(miles) as a predictor in a GLM framework, which naturally produces a concave relationship between miles and frequency. The choice of functional form should be guided by the data, with goodness-of-fit tests comparing alternative specifications.

It is important to distinguish between coverage types when modeling the mileage-frequency relationship. Liability and collision claim frequencies are strongly related to miles driven because these claims arise from driving activity. Comprehensive claims (theft, vandalism, weather damage, animal strikes) have a weaker relationship with mileage because many of these events can occur while the vehicle is parked. A well-designed PAYD program reflects these differences by varying the per-mile rate component by coverage type.

Pricing Framework

Pricing PAYD insurance requires decomposing the traditional annual premium into a fixed component (representing risk that does not vary with mileage) and a variable component (representing risk that increases with driving exposure). The fixed component covers comprehensive losses, a base level of theft and vandalism risk, and fixed administrative expenses. The variable component covers liability and collision losses and varies proportionally (or approximately proportionally) with miles driven.

The per-mile rate for the variable component is calculated by dividing the variable portion of the annual premium by the expected annual mileage, with adjustments for the non-linear mileage-frequency relationship. If the expected annual mileage for a rating class is 12,000 miles and the variable annual premium is $600, the per-mile rate is approximately $0.05. However, if the mileage-frequency relationship is concave, the per-mile rate should decrease at higher mileage levels, which can be implemented through a declining rate schedule or a capped maximum annual premium.

The total premium under PAYD consists of the fixed daily or monthly rate plus the per-mile rate times actual miles driven. Some programs also include a base monthly premium that covers a minimum number of miles, with additional charges for miles above the base. This structure ensures a minimum premium that covers the insurer's fixed costs while providing the mileage-based pricing signal.

Adverse Selection and Self-Selection

Adverse selection is a significant concern in voluntary PAYD programs. When PAYD is offered as an option alongside traditional annual policies, low-mileage drivers have a strong incentive to switch to PAYD (where they will pay less than the traditional annual premium), while high-mileage drivers have an incentive to remain on traditional policies (where they pay less per mile than they would under PAYD). This self-selection means that the PAYD population will have lower average mileage and lower average loss costs than the overall population.

From the insurer's perspective, this self-selection has several implications. The PAYD book should be priced based on the expected characteristics of the self-selected population, not the overall population. If the same company offers both PAYD and traditional policies, the traditional book will experience adverse selection as low-mileage drivers migrate out, potentially requiring rate increases for the remaining traditional policyholders. Over time, if PAYD adoption becomes widespread, the traditional book may shrink to a pool of high-mileage drivers, further intensifying the adverse selection effect.

Actuaries can model the expected adverse selection effect by analyzing the mileage distribution of the current book and estimating the switching behavior of policyholders at different mileage levels. Policyholders whose expected PAYD premium is lower than their current annual premium are likely to switch; those for whom PAYD would cost more are likely to stay. The magnitude of the adverse selection depends on the difference in premiums, the price sensitivity of policyholders, and how actively the company markets the PAYD option.

Behavioral Effects of Per-Mile Pricing

PAYD pricing creates a marginal cost for each additional mile driven, which may reduce total driving. Economic studies of mileage-based pricing consistently find that per-mile charges reduce vehicle miles traveled (VMT), with elasticity estimates suggesting that a 10% increase in the per-mile cost of driving reduces VMT by approximately 2 to 5%. This behavioral response creates a social benefit through reduced congestion, emissions, and accident risk, which has led some policymakers to advocate for mandatory PAYD pricing.

For actuaries, the behavioral response complicates the pricing analysis. If PAYD policyholders drive fewer miles than they would have under a traditional annual premium (because the per-mile charge makes them more conscious of their driving), the observed mileage in the PAYD population understates what the same drivers would have driven under traditional pricing. This means that simple comparisons of PAYD loss experience to traditional policy loss experience may overstate the risk reduction attributable to lower mileage and understate the behavioral effect of the pricing signal.

Regulatory and Market Considerations

Regulatory approval for PAYD programs requires demonstrating that the per-mile rates are actuarially justified. This means showing that mileage is a statistically significant predictor of loss, that the per-mile rate structure accurately reflects the mileage-loss relationship, and that the resulting rates are not unfairly discriminatory. Some regulators have expressed concern that PAYD could disproportionately affect lower-income drivers who rely on their vehicles for long commutes, though proponents argue that PAYD actually benefits many lower-income drivers who drive fewer miles than average.

Data verification is another regulatory concern. The credibility of mileage data depends on how it is collected. Self-reported mileage is subject to underreporting. Odometer readings at policy renewal provide annual mileage but not driving patterns within the year. Telematics devices and connected car data provide continuous, verified mileage data but raise privacy concerns. Regulators may require specific data collection methods or impose standards for mileage verification in PAYD programs.

Study Relevance

PAYD insurance concepts appear on CAS exams covering ratemaking and personal lines pricing, as well as on SOA exams covering general insurance. Candidates should understand the actuarial basis for mileage-based rating, the decomposition of premiums into fixed and variable components, the adverse selection dynamics of voluntary programs, and the behavioral effects of per-mile pricing. Practice problems might involve calculating per-mile rates from annual premium data, modeling the non-linear mileage-frequency relationship, or analyzing the expected adverse selection effect of introducing a PAYD option. These topics combine traditional ratemaking principles with behavioral economics and regulatory considerations, reflecting the multidisciplinary nature of modern actuarial practice.