600,000 Trips Exposed a Trendline Nobody Expected: The Faster You Drive, the More You’re on Your Phone

I ran the numbers, then I ran them again, and they kept saying the same thing: on limited-access highways, phone handling increases 12% for every 5 mph a driver exceeds the posted speed limit.^[1] Not from a survey or self-reporting but from 600,000 actual trips recorded by Cambridge Mobile Telematics insurance apps, analyzed by IIHS researchers Ian Reagan and Sam Monfort, published April 28, 2026.

Safety researchers had long assumed the opposite relationship. Drivers use phones in slow traffic, at stoplights, crawling through school zones, because the conventional logic holds that speed demands attention and fast equals focused. IIHS just ran a gyroscope against half a million GPS traces and found the assumption was backwards. On limited-access roads with free-flowing traffic, phone manipulation climbs with speed in a relationship so consistent it practically begs for a regression line.

Compound it out: a driver doing 75 in a 70 zone shows 12% more phone handling than someone at the limit, and at 80 that jumps to roughly 26%, while at 85 it hits about 40%. Someone doing 95 in a 70 zone is handling their phone 76% more than a driver at the posted speed.^[1] That is a 1.12ⁿ exponential curve, where n equals the number of 5 mph increments above the limit, and it holds across the dataset with a strength that made me check my arithmetic three times.

The relationship weakens on slower roads but never reverses. Arterials and connectors show a 3% increase per 5 mph increment. Roads with 45 or 50 mph limits produce a 3% larger effect than 25 or 30 mph roads. At 55 mph limits, that gap widens to 7%.^[1] Higher speed limits consistently amplify the correlation between speeding and phone use. Every cut of the data tells the same story.

"Until now, safety experts believed drivers used their cellphones most at slower speeds," IIHS President David Harkey said in the study's release. "But data from safe-driving apps show that, in free-flowing traffic, the opposite is true."^[1]

The Counting Problem Nobody Talks About

NHTSA reports 11,288 speed-related fatalities per year, representing 29% of all traffic deaths.^[2] Separately, the agency counts 3,275 distracted driving deaths, about 8% of fatal crashes.^[3] Two columns. Two totals. The IIHS data just proved those columns overlap in ways the coding system cannot capture.

When a driver doing 87 while scrolling Instagram rear-ends a stopped vehicle, which contributing factor absorbs the death? FARS, the federal crash database, relies on investigating officers to code contributing factors at the scene: alcohol gets a blood test, but distraction gets officer judgment and whatever the phone looks like in the wreckage. There is no mandatory post-crash phone forensic requirement in FARS data collection.^[2] The 3,275 distracted driving deaths are universally acknowledged as a severe undercount, and the IIHS study reveals exactly where those missing deaths hide: inside the speed column, attached to drivers who were doing both.

The Counterargument at Full Strength

IIHS itself flags this: correlation is not causation. Risk-seeking personalities may simply gravitate toward both behaviors simultaneously. A person who drives 90 and a person who texts while driving might be the same person not because speed produces phone use but because impulsivity, stress tolerance, and risk calibration push both needles at once. The telematics data cannot distinguish "bored at high speed" from "constitutionally incapable of leaving the phone alone regardless of speed." Personality is the confounding variable the study acknowledges but cannot control.

Additionally, the 12% figure applies specifically to limited-access roads in free-flowing traffic, and on arterials the effect drops to just 3%. On neighborhood streets at 25 mph, it may be negligible. Extrapolating the highway number to all driving conditions overstates the finding considerably.

Both objections are valid, but neither changes the policy implication: speed and distraction cluster in the same drivers, and treating them as independent risk categories understates the compound danger.

What This Means for You

If your insurer uses a telematics app, speeding and phone use now feed the same actuarial model, and there is published evidence that one predicts the other. Your premium reflects compound risk whether NHTSA counts it that way or not.

If you are evaluating vehicles with speed-limiting technology or phone-lockout systems, this study says those features solve the same problem from different angles. A car that limits top speed and a car that locks the screen above 10 mph are addressing the same behavioral cluster. Prioritize both.

If you commute on highways and consider yourself a "fast but attentive" driver: 600,000 trips of gyroscope data suggest that combination is rarer than you think.

Methodology

Phone handling is defined as gyroscope rotation consistent with picking up or manipulating a device while the screen is unlocked, per Cambridge Mobile Telematics' app instrumentation. Vehicle speed is matched to a commercial speed-limit database via GPS coordinates, with trips shorter than 18 minutes excluded along with time spent below 5 mph under the posted limit. Sample: approximately 600,000 trips recorded between July and October 2024. The 12% per 5 mph compounding calculation (1.12ⁿ) is The Crash Report's extrapolation from the per-increment rate reported by IIHS; the study itself reports the per-increment increase, not the compounded figure.

Limitations

Four states are excluded from the sample: California, New York, Alaska, and Hawaii. California alone represents the largest vehicle market in the United States, and its absence introduces significant geographic bias. The sample consists exclusively of insurance app users who voluntarily installed driving-behavior trackers, a population that almost certainly exhibits safer baseline behavior than unmonitored drivers. The true effect in the general population may be larger. Gyroscope-based phone detection cannot distinguish texting from adjusting GPS navigation from changing a song. FARS distraction coding relies on officer judgment rather than forensic phone analysis, making individual-crash-level cross-referencing between speed and distraction impossible with current data. The study covers a four-month window in 2024 and may not reflect seasonal or annual variation in driving behavior.

Source: IIHS telematics study (April 2026), NHTSA FARS 2014–2023, NHTSA distracted driving statistics. Insurance app data reflects a self-selected population. See methodology for caveats.