Data Analysis Shows Impact of Anti-Texting Laws

April 12, 2016
By Mark Austin

A few months ago, AT&T’s It Can Wait team asked our data scientists to look at something different from our normal work on network and business-process improvements:

Did we have any new ideas on how to study texting while driving?

The It Can Wait campaign shares a simple message: Keep your eyes on the road, not on your phone. And we did have an idea for them.

We looked at 3 months of anonymized data on our network. Through some algorithms and analysis, we were able to estimate the rates of texting while driving across the United States.

The real world creates a lot of variables. So we’re not announcing rates for individual cities. There’s no need to create a fuss when we might call the numbers directionally accurate, rather than pinpoint. But, after some checking and re-checking, our data science team is confident of this general statement:

States that have statewide anti-texting laws have lower rates of texting while driving – at a statistically significant level.

We believe that the 4 states without a full statewide ban have a roughly 17 percent higher rate of texting while driving than the 46 states with statewide bans.

Here is a closer technical look at what we did:

  • Privacy first. We didn’t create any new data, nor did we share any data outside of AT&T. Our data scientists looked at an anonymized set of routine network information, and aggregated it to the size of a metropolitan area.
  • Specifically, we looked at outgoing text messages, and we used a cell-tower algorithm to figure out which ones were sent from moving vehicles.
  • We studied commutes within US Census metropolitan areas. We identified “commutes” as trips taken between an anonymous phone’s 2 most frequented tower locations. (The actual locations were not relevant beyond verifying that the towers were in a particular metro area.)

Here’s why we did it this way:

  • We could only know that a mobile device was on the move, not whether it was being used by a driver or passenger. So our solution was to “weight” the rate in each metro area. We stuck to commutes in metro areas because the US Census collects passenger data specifically about metro-area commutes. In most areas, roughly 80 percent of commutes are solo – a car with no passenger. But in the greater New York area, for instance, the figure is 50 percent. We adjusted every metro area for this bias to get a truer picture of texting while driving, not just texting while moving.

We think this study is significant because it shows what people as a group are actually doing, not what they say they are doing. And remember, this is just texting – not other smartphone driving distractions. But it’s a reasonable indicator.

Should the 4 states join the rest of the states by passing comprehensive texting bans? We’re happy to add to the research.

We’ll share our insights with our fellow members of Together for Safer Roads, a coalition that addresses global road safety. As a single study, we hope this work will help raise awareness. It could be replicated in the future as legislation changes, or adjusted with new ideas.

The study of anonymous and aggregated data from mobile phones has great potential for good. In California, for instance, AT&T is helping on a project to determine how mobile data may save taxpayers literally billions of dollars. With aggregated traffic data from phones, the state may not need to replace worn-out, in-pavement traffic sensors to conduct congestion studies.

It’s something to think about. But not to text about while driving.

Mark Austin is vice president for data insights, Big Data at AT&T.

Filter by:

  • Most Popular
  • Author
  • Tags

Archives

Link to archives