Thursday, 25 October 2012

Texting while driving

Texting while driving is dangerous and increases your chance of accidents. Bans on texting while driving seem to increase accident rates.

The Insurance Institute for Highway Safety says that 3 of every 4 states that have enacted a ban on texting while driving have seen crashes actually go up rather than down.
It's hard to pin down exactly why this is the case, but experts believe it is a result of people trying to avoid getting caught in states with stiff penalties. Folks trying to keep their phones out of view will often hold the phone much lower, below the wheel perhaps, in order to keep it out of view. That means the driver's eyes are looking down and away from the road.

So, here's a potential difference-in-difference study in this, the "thinking about honours projects for next year" time of year.

New Zealand banned texting while driving a few years ago. If we can get data on age-by-age participation in texting (presumably more prevalent among young people than among older people), we'd expect that the ban would have no effect on older cohorts' rates of single-vehicle accidents, that accident rates among youths will depend on the elasticity of texting risk-taking with respect to the law (could go either way - if the number of texting kids drops enough to offset the increased risk-taking by those still texting, accident rates drop; otherwise, they increase), and that multivehicle accident rates will change proportionately to the types of drivers involved (no change to multivehicle accidents where elderly drivers smash into each other, other cross-effects depend on the elasticity found in the single-vehicle accident rates).

Data we need:
  • Age-by-age texting participation rates (don't know if any proxy exists)
  • Age-by-age accident rates  (near certain this exists)
  • Accidents broken down by ages of all drivers involved and single- or multi-vehicle status
  • Checking whether other age-specific interventions coincided with the texting law change:
    • The zero-alcohol limit for young drivers (which itself would make for a great diff-in-diff study not only on accident rates but also on crime rates)
    • The 'Crusher Collins' crackdown on 'boy racers'
Hints on data warmly accepted.

Update: the original study is here.


  1. I suggest adding in a severity of accidents variable as well. The total number of fender benders may well go down (fewer text and drivers) while the severity of related accidents goes up (for the reasons you listed)

  2. I'll have to check whether the by-age data on accidents includes severity. Thanks!

  3. Relevant:

  4. Perhaps a dummy variable - "can you text without looking at your phone?". This is probably correlated with both the chance of accident and with the age-by-age participation rates.

  5. Absent individual-level data, I might be forced to use US state-level data. I'd be surprised if that weren't already done though.