Driving fatalities after 9/11: a hidden cost of terrorism

Abstract

We show that the public's response to terrorist threats can have unintended consequences that rival the attacks themselves in severity. Driving fatalities increased significantly after the 11 September 2001 terrorist attacks, events that prompted many travellers to substitute road transportation for safer air transportation. After controlling for time trends, weather, road conditions and other factors, we find that travellers’ response to 9/11 resulted in 327 driving deaths per month in late 2001. Moreover, while the effect of 9/11 weakened over time, as many as 2300 driving deaths may be attributable to the attacks.

Acknowledgements

We thank Michael Waldman, James E. Blalock, Kosali I. Simon and an anonymous reviewer for extensive comments and discussions.

Notes

¹ For example, in a discussion of whether infants on airplanes should be placed in child seats (New York Times 4 August 2004; also see Windle and Dresner, 1991), the Federal Aviation Administration was concerned about whether overall risk to infants would increase if families substituted unsafe driving rather than paying airfare for children under the age of two. Hahn (1996) uses the same logic to argue against draconian security measures in airports that make travellers switch from air travel to surface transportation.

² An interesting contrast is provided by Fischoff et al. (2003) which reports results of a November 2001 survey showing systematic underassessment of the risk of being involved in a terrorist attack. It attributes this underassessment to ‘motivational biases (e.g. the desire to feel more secure) or cognitive ones (e.g. not realizing how much easier it is to see one's own precautionary measures, compared to others)’. Kunreuther and Michel-Kerjan (2004) explain that the corporate demand for terrorism risk insurance is in part due to underassessment of risk.

³ We were conservative in determining which vehicle categories to include. We excluded some vehicles that FARS categorizes as passenger vehicles (large utilities, compact pick-up trucks, standard pick-up trucks, etc.). When we include these additional passenger-vehicle categories, our results remain largely unchanged.

⁴ We obtained data on these prices from the Bureau of Labor Statistics.

⁵ We obtained unemployment rate data from the Bureau of Labor Statistics and income data from the Bureau of Economic Analysis.

⁶ One might be concerned that changes in state regulations or other regional conditions may have influenced fatalities. For example, stricter drunk-driving laws may have reduced fatalities, while higher speed limits may have increased fatalities. However, because the above results use national data, the net effect on fatalities of any unobserved state-level changes is likely to be small and should be captured by the year fixed effects. Nonetheless, to verify that state-level changes are not biasing our results, we disaggregate our data to create a noncommercial fatality count for each state in each month. We then estimate the effect of 9/11 on these fatality counts and include state-year fixed effects to control for unobserved changes in state conditions. The effect of 9/11 increases and suggests that any bias from unobserved state-specific factors is negative.

⁷ A simple demonstration clarifies the point. If we redefine the 9/11 dummy to take a value of one only for the last 3 months of 2001 and then re-estimate Equation 1, we obtain identical results, except that the 2002 and 2003 fixed effects each shift up by exactly 327.469, the coefficient on the 9/11 dummy.

⁸ For one possible explanation, see Steklov and Goldstein (2004) which shows that terrorist attacks in Israel produce a 2–3 day slump in nonfatal road accident rates, followed by a 35% increase in fatal accident rates beginning 3 days after the attack. The article speculates that this pattern could partly be attributable to attack-induced stress.

⁹ We calculate the fraction of total fatalities attributable to the increase in VMT, and then multiply this by the number of additional noncommercial fatalities per month post 9/11: (0.048/0.085) × 327.469=184.924.