U.S. fatalities from traffic accidents rose 7.2% last year to 35,092—the largest increase in 50 years—and distracted driving played a role in 10% of those deaths, according National Highway Traffic Safety Administration figures released last month. NHTSA found that fatalities from “distraction-affected” crashes increased 8.8% to 3,477 from 3,197 for that period.

Given how many drivers seem to be on the phone, applying makeup, fiddling with radios, or texting while merging lanes, that figure might actually seem low.

Everquote’s research, based on a new survey of 2,300 drivers, stresses that cognitive dissonance. Everquote combined the survey responses with data gleaned from 35,000 U.S. drivers using its Everdrive app on smartphones, producing some thought-provoking conclusions.

For example, nearly every one (96%) of the survey respondents said he or she is a safe driver, but 56% of the same group admitted to using the phone while driving.

Data from the Everdrive app, on the other hand, revealed 96% of all drivers used their phones at least once in the past 30 days, and drivers averaged about one call per trip. For every 11 miles driven, the “average” driver is on the phone for 0.4 miles.

The survey does not distinguish between use of the phone for talking or for texting, which is arguably more dangerous. An Everquote spokesman said that, for privacy reasons, it has does not access that sort of data. What it does know is whether drivers unlock their phones and if the phone is moving consistent with a moving vehicle. Another wrinkle: Conversations will not register if the driver uses speaker phone or bluetooth and so does not actually pick up the handset.

Excessive speed, as always, is another big factor in traffic fatalities. NHTSA data found that deaths in crashes involving speeding rose 3% last year to 9,557 from 9,283.

But 42% of the surveyed drivers said they don’t consider going 10 mph over the speed limit to be speeding. Another 10% said they don’t think a 20 mph increase is speeding. Meanwhile, the app data showed that drivers speed at least 10 mph more than half of the time. National data shows that even a 10 mph speed increase ups the risk of a crash by 9.1%.

For more on self-driving cars, watch:

Unsurprisingly, most (81%) of drivers surveyed felt they are better, safer drivers than any self-driving car from Tesla (TSLA), Google (GOOG), or any other company could be. Meanwhile, 2015 research from McKinsey posits that self-driving cars will reduce crashes up to 90% over those driven in the traditional sense (by human drivers).

The scary thing is things may get worse, given that most automakers seem to be packing more “infotainment” gear in each new car produced these days.

Note: This story was updated with more information on what the survey measures and does not measure.