An Interactive Tool to Compare Traffic Risks Traffic STATS

1
An Interactive Tool to Compare Traffic Risks
Traffic STATS

• Paul S. Fischbeck, David Gerard, Barbara Gengler,
  and Randy S. Weinberg
  
• Center for the Study Improvement of Regulation
• Carnegie Mellon University
• INFORMS Annual Meeting
• Pittsburgh, PA
• 6 November 2006

2
Why Provide Travel Risk Information?

• Motor vehicles involved in 40,000 fatalities
  each year (including drivers, passengers,
  pedestrians)
  
• Accounts for 1 in every 55 deaths in the U.S.
• Leading cause of death for every age from 3 to 33
  
  
• Travel risks are not the same
• For all ages
• For all times
• In all places
• In all types of vehicles
• How should we think about risk?

3
Media Portrayal of Risk

• When are young drivers most at risk?
• October 25, 2006
• For teens, driving the fast lane to independence.
  But now, a word of caution from AAA commuting
  home from school between 3 and 5p.m. may be just
  as harmful to your 16- or 17- year-old's health
  as cruising during high-risk weekend hours. AAA
  says that between 2002 and 2005, young drivers
  were involved in 1,100 fatal crashes during
  weekday periods, almost as many as weekends, a
  little more than 1,200.

4
Media Portrayal of Risk
People driving or riding in a sport utility
vehicle in 2003 were nearly 11 percent more like
ly to die in an accident than people in cars
The traffic safety agency reported last week th
at there were 16.42 deaths of S.U.V. occupants i
n accidents last year for every 100,000
registered S.U.V.'s. The figure for passenger ca
rs was 14.85 deaths for each 100,000 registered
5
Fatalities ? Risk

• It is not possible to understand underlying risks
  by simply looking at the number of fatalities
  
• 18,819 fatalities for vehicle occupants in
  passenger cars
  
• 3,779 motorcycle fatalities
• Risk depends on exposure how much a mode is
  used.
  
• Passenger cars are driven much more than
  motorcycles
  
• Passenger car risk is 1.05 per 100 million
  passenger miles traveled
  
• Motorcycle risk is 32.61. Over 30 times riskier!
  
  
• Possible risk measures
• Deaths per 100,000 registered vehicle
• deaths per 100 million passenger miles
• deaths per 100 million trips
• deaths per 100 million minutes of travel

6
Hot Topic Example

• Which are riskier SUVs or Cars?
• Number of fatalities
• Fatalities per registered vehicle
• Fatalities per vehicle mile
• Fatalities per passenger mile
• Which is easiest to support with data?
• Which is the best measure of the true risk?
• Which is most useful for policy and decision
  making?

7
Risk Measures (2001)
Source NHTS and FARS 2001
8
Why the Differences?
Statistical averages for 2001
Source NHTS 2001

• The greater exposure in SUVs makes the
  denominator larger and the relative risk smaller.

9
Risks Relative to Cars
10
Construction of Risk Measures

• Risk characterized by ratios
• Numerator Outcome (fatalities)
• Denominator Exposure (miles, trips, minutes)
• Two underlying government databases for travel
  risks
  
• FARS (fatalities)
• NHTS (exposure)
• Constructing the ratios requires some effort
• Identify common attributes in the two datasets
• Calculate confidence intervals
• Compare risks

11
Fatality Analysis Reporting System (FARS)

• Contains data on all U.S. vehicle crashes that
  result in one or more fatalities.
  
• Crash must involve a motor vehicle on a public
  traffic way that results in the death of a person
  (either an occupant of a vehicle or a
  non-motorist) within 30 days of the crash.
• Contains detailed descriptions of each fatal
  crash (more than 100 elements characterize the
  crash, the vehicles, and the people involved)
  
• Updated annually
• FARS is a complete count of all fatalities (no
  uncertainty)
  
• TREADS contains select FARS data from 1999-2004

12
National Household Travel Survey (NHTS)

• NHTS surveys the nations inventory of daily and
  long-distance travel
  
• Survey is completed once every five to seven
  years
  
• Sample of U.S. households used to provide
  national estimates of trips and miles by travel
  mode, trip purpose, and a host of household
  attributes
• Contains demographic characteristics of
  households, people, vehicles, and detailed daily
  and long-distance travel information for all
  purposes by all modes
• Survey data is by definition a sample and there
  is uncertainty in these national estimates

13
Uncertainty and Confidence Intervals

• Risk estimated by taking the number of fatalities
  and dividing by estimated number of miles, trips,
  or minutes.
  
• The actual number of miles traveled is unknown
  it is derived from the NHTS sample and
  therefore both the denominator and the related
  risk estimate are uncertain.
• Using statistical techniques based on replicate
  weights (the number of people in the country that
  each sample person represents), a confidence
  interval can be estimated.
• The width of the interval depends on the NHTS
  sample size, but generally larger numbers of
  fatalities are associated with tighter confidence
  intervals.
• If the interval includes zero (0), then the
  results are not statistically significant.
  
• Overlapping intervals do not necessarily imply
  insignificant differences
  

14
Traffic STATS STAtistics for Travel Safety

• Joint project
• Carnegie Mellon University
• AAA Foundation for Traffic Safety
• Capabilities
• Easy access to underlying data
• Interactive, user-friendly interface
• On-the-fly risk calculation for millions of
  combinations of input variables
  
• Confidence intervals calculated when necessary
• Layout of the website
• Online tutorial
• FAQ

15
Basic Web Layout
16
Millions of Comparisons

• Transportation Mode Categories
• All transportation modes
• Personally owned vehicles
• Car
• Van
• SUV
• Pickup truck
• Motorcycle
• School bus

• Risk Comparisons Categories
• Age
• Day of Week
• Gender
• Hour of Day
• Month/Quarter
• Person Type
• Region
• Transportation Mode

• Other bus
• Local public
• City-to-city
• Charter/tour
• Walking
• Bicycling

17
Vehicle Type Risk across Regions
18
Seasonal Risks across Regions
19
Rollover Risk by Age SUV vs. Car
20
Overall Risk by Age SUV vs. Car
21
Pedestrian Risk and Age
22
Driving Risk and Time of Day
23
Pedestrian Risk and Time of Day
24
Driving Risk and Age
25
Applications

• Results from the Traffic STATS searches is easily
  exported into Excel for detailed statistical
  analyses
  
• Travel safety research
• Risk communication
• Press
• Education
• Setting the policy agenda
• General public

26
Acknowledgements

• AAA Foundation for Traffic Safety
• Carnegie Mellon University
• Center for the Study Improvement of Regulation
• Information Systems Program
• Barbara Gengler at Multidimentionality, LLC

Contact Information
Paul Fischbeck 412-268-3240 fischbeck_at_cmu.edu
David Gerard 412-268-1273 dgerard_at_cmu.edu
27
(No Transcript)