For almost 20 years, it has been a wide-held belief that talking on a cellphone while driving is dangerous and leads to more accidents. However, new research from Carnegie Mellon University and the London School of Economics and Political Science suggests that talking on a cellphone while driving does not increase crash risk.
Published in the American Economic Journal: Economic Policy, the study uses data from a major cellphone provider and accident reports to contradict previous findings that connected cellphone use to increased crash risk. Such findings include the influential 1997 paper in the New England Journal of Medicine, which concluded that cellphone use by drivers increased crash risk by a factor of 4.3 — effectively equating its danger to that of illicit levels of alcohol. The findings also raise doubts about the traditional cost-benefit analyses used by states that have, or are, implementing cellphone-driving bans as a way to promote safety.
“Using a cellphone while driving may be distracting, but it does not lead to higher crash risk in the setting we examined,” said Saurabh Bhargava, assistant professor of social and decision sciences in CMU’s Dietrich College of Humanities and Social Sciences. “While our findings may strike many as counterintuitive, our results are precise enough to statistically call into question the effects typically found in the academic literature. Our study differs from most prior work in that it leverages a naturally occurring experiment in a real-world context.”
For the study, Bhargava and the London School of Economics and Political Science’s Vikram S. Pathania examined calling and crash data from 2002 to 2005, a period when most cellphone carriers offered pricing plans with free calls on weekdays after 9 p.m. Identifying drivers as those whose cellphone calls were routed through multiple cellular towers, they first showed that drivers increased call volume by more than 7 percent at 9 p.m. They then compared the relative crash rate before and after 9 p.m. using data on approximately 8 million crashes across nine states and all fatal crashes across the nation. They found that the increased cellphone use by drivers at 9 p.m. had no corresponding effect on crash rates.
Additionally, the researchers analyzed the effects of legislation banning cellphone use, enacted in several states, and similarly found that the legislation had no effect on the crash rate.
“One thought is that drivers may compensate for the distraction of cellphone use by selectively deciding when to make a call or consciously driving more carefully during a call,” Bhargava said. “This is one of a few explanations that could explain why laboratory studies have shown different results. The implications for policymakers considering bans depend on what is actually driving this lack of an effect. For example, if drivers do compensate for distraction, then penalizing cellphone use as a secondary rather than a primary offense could make sense. In the least, this study and others like it, suggest we should revisit the presumption that talking on a cellphone while driving is as dangerous as widely perceived.”
Pathania, a fellow in the London School of Economics Managerial Economics and Strategy group, added a cautionary note. “Our study focused solely on talking on one’s cellphone. We did not, for example, analyze the effects of texting or Internet browsing, which has become much more popular in recent years. It is certainly possible that these activities pose a real hazard.”
For more information, visit http://www.cmu.edu/dietrich/sds/people/faculty/saurabh-bhargava.html.
who says it wasn’t. a passenger on the phone?
Not having read the original study, I can’t be sure that the problems I see with it are actually problems with the study or problems with the way it was interpreted in the Science Blog. From reading the Blog, however, I see a few problems with the study methods for cell phone (CP) user crash data.
1. The proxy used to identify calls from drivers vs non-drivers is “calls that were routed through multiple cellular towers.”
Driving a car while using a CP doesn’t necessarily mean that the call will be routed through multiple towers. So while the proxy may correctly identify some drivers, the rest are lumped in with non-drivers. The use of multiple cellular towers may, in fact, identify only longer-distance drivers.
2. The study identifies a spike in use at 9 p.m. “…drivers increased call volume by more than 7 percent at 9 p.m.”
a. If free calls encouraged drivers to make more calls after 9 p.m., non-drivers would have been similarly encouraged. Right? Yet the study does not mention an increase in calls for the “non-driver” group. That calls into question the validity of the method for defining the groups.
b. If the free pricing plans were meant to attract more customers to make use of the excess after-business capacity (a likely assumption), the non-driving group would, in fact, also increase their calling rate, CP companies would have had to increase their capacity to accomodate calls that did not increase revenue. Did that happen?
c. Long-distance drivers may behave differently from other drivers after 9 p.m. They might call home to say update their arrival time. Or they could use phone conversation to help keep awake. So again the question of whether the “driver” group represents the entire CP using driving population or just the long-distance portion of it.
3. The analysis of accident rates for those states that prohibited cell phones while driving “… found that the legislation had no effect on the crash rate.”
There is no mention whether or not the study determined if the legislation had any effect on the use of CP while driving. If there’s no effect on CP use, then the legislation would be irrelevant to the study.
4. The study does not explain the difference in night time traffic accidents for this study to historic night time traffic accident data.
Most studies show an increase in night time traffic accidents and fatalities despite a reduction in miles driven. The data on 8 million crashes across 9 states used for this study apparently do not.
Discussion: Using generalized accident data that appear to differ from other historic accident data to compare with other generalized data from a population that may not be the one intended will not deliver a useful conclusion.
Conclusion: GIGO