Svenson's 1981 study found the overwhelming majority of drivers believe they're safer and more skilled than their peers. That self-perception gap is exactly why self-reported driving data doesn't work.
In 1981, psychologist Ola Svenson asked 161 students in Sweden and the United States to rate their own driving against their peers. Mathematically, about half of any group has to be below the median. That’s not what people reported.
93% of the U.S. sample and 69% of the Swedish sample placed themselves in the top 50% for driving skill. 88% of the U.S. sample and 77% of the Swedish sample placed themselves in the top 50% for safety. The paper’s title asks the question directly: “Are We All Less Risky and More Skillful Than Our Fellow Drivers?” (Svenson, 1981, Acta Psychologica 47, 143–148).
This isn’t a one-off finding. A 1986 replication by McCormick, Walkey, and Green found 80% of 178 participants rated themselves above average across eight separate driving-skill dimensions. A more recent commercial survey found 36% of drivers still believed they were an above-average driver while texting or emailing behind the wheel — rating themselves safe in the exact moment behavioral data says otherwise.
Psychologists call the general pattern the better-than-average effect, and driving is one of its most consistent domains, for reasons that are specific to how driving feedback works:
If the overwhelming majority of drivers rate themselves above average, self-assessment isn’t just imprecise as an input to pricing or coaching — it’s systematically biased in one direction.
This is the actual argument for objective, device-based measurement in a telematics program, stated plainly: it’s not that self-report is slightly noisy, it’s that it’s structurally optimistic in exactly the population a safety or pricing program most needs to identify. A driver who self-selects into “I’m already safe, I don’t need to change anything” is precisely the driver a hard-braking or phone-distraction event, measured on-device and reported back per trip, is meant to reach.
It also sets a bar for what the measurement itself has to get right. If the entire premise is “don’t trust self-report, trust the sensor,” then the sensor has to be more reliable than the bias it’s replacing — attributing hard-braking events to the driver and not the road, distinguishing driver phone use from passenger phone use, and doing both consistently across the actual device mix a fleet runs, not just flagship phones in a lab. That’s the standard Sampark is built and tested against, and it’s why Bodha Drive shows drivers a per-trip breakdown instead of one averaged score — a driver correcting a specific, dated phone-distraction event learns something self-assessment never could.