Mobile broadband access—used as an indicator of the likelihood that drivers may be using mobile devices while driving—appears to have no impact on collision frequency, according to a recent insurance industry research report.

The Auto Loss Cost Trends Report, a product of research undertaken jointly by The Casualty Actuarial Society, Society of Actuaries and Property Casualty Insurers Association of America, was sparked by an observed increase in the frequency and severity of private passenger auto losses in recent years. After more than 25 years of declining auto crashes, personal auto insurance carriers began to notice an uptick in property damage liability and collision frequency in 2013, the report notes, pointing to popular theories like better safety awareness, technology and enforcement as explanations for the favorable prior trend.

But what explained the reversal?

While mobile device use surprisingly doesn’t appear to be the main driver of several subsequent years of frequency spikes, the research seems to point to a simple explanation: increased road congestion.

Using publicly available data from the Federal Highway Administration, Bureau of Labor Statistics, the Census Bureau and other sources, the analysis group searched for explanatory variables among possibilities that included:

• Percentage of population with access to mobile broadband (variable name: Mobile Broadband Percent).
• Gas price indicators—specifically, average gas price in dollars divided by average hourly wage in dollars (Gas Price vs. Wage).
• Total transportation dollars spent on capital projects per vehicle miles traveled (Capital Outlay per VMT).
• Number of lawyers in the state per one million people (Lawyers per 1 Million Capita).
• Total DUIs per driver.
• Ratio of licensed drivers to total lane miles (Drivers per Lane Mile).
• Average commute time in minutes in urban areas and in rural areas (Urban Average Commute Time, Rural Average Commute Time).
• Percent of the vehicle miles traveled in an urban area or in a rural area (Urban VMT Percent, Rural VMT Percent).

Searching for relationships between these state-level variables and quarterly claim frequencies also captured by state (from fourth-quarter 2011 through fourth-quarter 2015), the groups found that the variables showing the strongest positive relationships with collision frequency were those related to congestion—Drivers per Lane Mile, Urban and Rural Average Commute Times, Urban VMT.

The weakest variables were Mobile Broadband Percent, Gas Price vs. Wage and Capital Outlay per VMT, with variables measuring unemployment, lawyer concentrations and DUIs falling in the middle of a variable ranking based on importance. In fact, DUIs appear to be negatively correlated with collision frequency, the report notes.

The report suggests, however, that the weak relationship between mobile broadband access and collision frequency isn’t entirely conclusive evidence that distracted driving was not a factor behind upticks in crash rates. Instead, it is “likely [that] we need to find a better proxy for distracted driving” than mobile broadband access, the report states.

Finding another proxy might be a challenge, according to information released separately by Cambridge Mobile Telematics last week. Only 11 states have “mobile-phone distraction” as an accident reporting option for police, and only 27 states can note “distraction” generally, CMT said in an announcement about results of a recent driver survey comparing motorist opinions of DUI and distracted driving.

According to the CMT survey of 700 U.S. drivers, 75 percent said that they see other drivers on phones every single day, and 63 percent are more afraid of distracted drivers than drunk drivers.

As for the CAS, SOA, PCI report, other findings from the analysis included these:

• Massachusetts, Michigan, Maryland and New York have the highest average collision frequency.
• South Dakota, Idaho, Wyoming and Montana have the lowest average collision frequency.
• The type of tort system (no-fault, optional no-fault or tort) does not appear to significantly impact the expected collision frequency. System type, however, may have an impact on the variance of the frequency around the expected value.

The latter finding is subject to further investigation, Richard Gorvett, staff actuary for the CAS, explained in an email responding to a Carrier Management request for more information. He noted that the preliminary analysis shows that tort states have just slightly lower expected or average frequencies than non-tort states. They also have somewhat lower variability around the averages—a relationship that does not seem intuitively predictable and calls for further analysis.

Separately, insurers have been reporting better news related to personal auto loss trends. The Hartford, for example, reported flat frequency and low single-digit severity trends during its fourth-quarter earnings call, partially explaining improved profitability for the personal auto line. Allstate reported declining frequency trends for bodily injury and property damage coverages, with frequency coming in below elevated levels experienced in 2015 and 2016 across the United States. Both companies also attributed profit improvements in the line to pricing and other actions.

Christopher Swift, chief executive officer of The Hartford, speaking at the Bank of America Merrill Lynch Insurance Conference last week, said, “Probably, my biggest pleasure in talking about [2017] comes from the dramatic improvement in personal lines.” He was referring to improving company fundamentals following frequency spikes that surprised the industry in 2016—and The Hartford, in particular, which had to report adverse development in that year. In addition to rate and underwriting actions, The Hartford pared back its agency force, he said. “We grew too fast, and we just needed to course correct. Compounded with…the rise and spike in frequency and severity, [it] just put pressure on us. So we knew it was fixable,” he said, noting that he is not totally surprised by the rapid turnaround in results.

Swift was one of several CEOs to sound the alarm on distracted driving in late 2016, when the company results were impacted by the overall trends that the CAS, SOA and PCI are investigating now. At the time, even as Swift described the internal plans for improvement, he delivered strongly worded messages to the driving public. “We can teach ourselves to put down devices…Put them down, get there safely, and then pick back up your activity. Teach everyone to do that,” he said during a December 2015 investor conference. “When we talk about speed of accidents, it’s not because everyone is driving faster. It’s because there’s less reaction time for braking…That’s classic distracted driving,” he said. (See related article, “Too Much Tech, More Lawyers Fuel Auto Insurance Woes: Hartford Execs.”)

“Educate your friends, your family, your children on distracted driving. It is a real issue that’s causing real pain for society in general,” he said.

Describing current frequency trends as being down and flattening—but still elevated from where they were three years ago—The Hartford CEO could only guess at the reasons when asked for his take at the Bank of America Merrill Lynch conference last week. “It could be safety. It could be less distracted driving. There [have] been more campaigns from an awareness side…We’ve had, maybe, road conditions improved a little bit. But it’s just hard for us to look at a dataset over one year” to pinpoint “the root cause of any improvement in a one-year time horizon.”

The multiyear analysis in the CAS, SOA, PCI Auto Loss Cost Trends Report (January 2018) is the first of several such reports planned by the groups. While the current report deals with collision, future reports will examine other auto insurance coverages.