ADAS Creates a Frequency-Severity Paradox for Auto Insurers: Fewer Crashes, Costlier Claims

Two March 2026 datasets landed within days of each other, and read together they describe one problem. The Highway Loss Data Institute measured how much collision frequency advanced driver assistance systems remove. CCC Intelligent Solutions measured the cost layer that arrives with them. The combined picture is not a trend adjustment. It is a change in the relationship between frequency and severity that standard personal auto pricing models are not built to hold.

Key Takeaways

• 39% fewer property damage liability claims on Mazda's most comprehensive ADAS bundle, against 13% for a basic front-AEB and forward-collision-warning pair, in HLDI's study of 2015 to 2023 model years.
• 28.3% of repairable estimates carried an ADAS calibration in 2025, up from 21.8% a year earlier, at an average fee of $485.56, alongside a record 23.1% total loss rate.
• 51.5% of calibrations arrive as supplements rather than on the initial estimate, adding a cost layer to auto physical damage development that pre-2023 triangles do not contain.
• $42 per vehicle, roughly 10% of pure premium, separates the independent from the correlated pure premium in CCC-era arithmetic: multiplying a frequency trend by a severity trend double counts a compositional shift.
• 12.8 years is the average age of the US light vehicle fleet, so ADAS and non-ADAS vehicles sit inside the same rate filing for most of a decade.

What the Two March Datasets Measured

HLDI, the data arm of the Insurance Institute for Highway Safety, examined six feature bundles and four standalone systems across 2015 to 2023 Mazda vehicles. The informative result is the gradient, not the headline. A basic bundle of front automatic emergency braking and forward collision warning was associated with 13% fewer property damage liability claims and 9% fewer bodily injury liability claims.

Each addition moved the number. Blind spot monitoring with rear cross-traffic alert took off nearly 10% more PDL claims and 13% more BIL claims. The comprehensive bundle, layering pedestrian detection, adaptive cruise control, high-beam assist, lane departure warning and prevention, rear AEB, and driver attention alerting, reached 39% on PDL and 21% on BIL. HLDI flagged that the 21% bodily injury figure was not statistically significant at conventional levels, reflecting thinner BIL claim counts.

HLDI chief insurance operations officer Matt Moore named the mechanism: crash avoidance works at low speeds. Front AEB addresses rear-end crashes, rear AEB removes parking lot incidents, lane departure prevention stops single-vehicle run-off-road events. Each system targets the high-frequency, low-severity end of the distribution.

CCC's Crash Course 2026 report supplies the other half. ADAS calibrations appeared on 28.3% of repairable estimates in 2025, up from 21.8%, and reached 35.6% in DRP shops by the third quarter. Calibration fees averaged $485.56; diagnostic scans averaged $149.10 and appeared on 87.7% of collision estimates. As our CCC Crash Course 2026 analysis set out, the 1.7% headline repair cost increase sits above divergent sub-trends: newer vehicles averaged $5,721 against $3,682 for older ones.

Why the Two Trends Cannot Simply Be Multiplied

Personal auto pricing treats frequency and severity as independent random variables. Pure premium is frequency times severity, and each side is trended separately. That independence assumption is embedded in most GLM frameworks and in the trend exhibits that support rate filings in every state.

ADAS breaks it. The same technology that removes claims also raises the cost of the claims that remain, so the two components are negatively correlated at the vehicle level through a single causal channel. The consequence is that a multiplicative model overstates loss costs for ADAS-heavy cohorts.

Scenario	Frequency	Severity	Pure Premium
Pre-ADAS baseline	10.0 claims per 100	$4,200	$420
ADAS: independent model	7.0 claims per 100	$5,500	$385
ADAS: correlated model	7.0 claims per 100	$4,900	$343

In the independent model, the $1,300 severity increase applies to every surviving claim. In the correlated model the increase is $700, because the claims ADAS eliminated were the cheapest ones; the average rises mechanically rather than because comparable repairs cost more. The $42 gap is roughly 10% of pure premium.

The working split is between true repair cost inflation on claims that would have occurred anyway and compositional inflation caused by removing cheap claims from the pool. Only the first should compound forward. The second is a one-time level shift in the severity base.

Whether a carrier can make that split depends on data many do not hold. Progressive's 21 million telematics-connected policyholders, growing at a 28% CAGR since 2018, can be resolved to individual vehicle configurations through OEM feeds. A carrier rating on model year as a proxy blends ADAS and non-ADAS vehicles inside one class and cannot separate the two severity components at all. The gap is widest in bodily injury, where CCC put severity up 10.3% year over year and 32% over four years, now 52.4% of liability dollars paid, against social inflation running on its own track.

The Segmentation Fix Collides With the Rate Filing

The obvious response, segmenting trend by ADAS equipment level, is harder to file than to calculate. Several states restrict vehicle-level rating variables that regulators read as socioeconomic proxies, and comprehensive ADAS packages sit disproportionately on more expensive vehicles. HLDI's finding that nearly all bundles were associated with lower total property damage losses is the statistical foundation for the variable, but the filing still has to carry a disparate impact analysis showing the correlation is incidental to the risk effect.

Some carriers route around the problem by rating behavior instead of equipment. A telematics program that observes fewer hard braking events and fewer lane departures captures much of the same segmentation, and it survives restrictive rate regulation more easily, which is one reason model validation practice in state filings has moved toward behavioral variables. It is a correlate of the equipment rather than the equipment itself.

The second constraint is time. With the fleet averaging 12.8 years, and S&P Global Mobility projecting eight ADAS systems in half or more of registered US vehicles by 2029 against three by 2027, the book runs as two populations for most of a decade. The newer cohort produces fewer claims at higher cost per claim. The older cohort keeps its historical frequency while carrying a 45.3% total loss rate at 13 years and older, plus tariff-driven parts inflation.

The compositional effect is also self-limiting. As penetration approaches saturation there are progressively fewer cheap non-ADAS claims left to remove, so the mechanical severity lift fades even while repair cost inflation continues. A carrier that books the compositional shift as ongoing severity trend will keep projecting it forward at exactly the point it stops arriving.