# What Exacerbates Injury and Harm in Car SUV Collisions?

Size: px
Start display at page:

## Transcription

4 Table 2. Tabulation of Harm by Different Two-Vehicle Crash Types Collisions N Percent % Mean car-harm Mean SUV-harm Mean harm in crash Standard deviation of harm Mean harm per person Standard deviation of harm per person Car car 1,229, ,798 83, ,580 28,150 93,844 Car SUV 402, ,932 39, , ,605 40, ,534 SUV SUV 22, , , ,939 49, ,448 harm is strongly positively skewed, partly due to the very high costs of fatalities. For dummy variables, the marginal effects for x m equal the difference in harm per person with a discrete change in variable x m from 0 to 1 with all other variables at their modal values. The equation follows: k y xm =1 y xm =0 = exp 0 + i x i + m k exp 0 + i x i, where i m 6 Marginal effects for the only interval independent variable in this study, the number of passengers x m, are calculated by the following equation: k * dy/dx m = m exp 0 + i x i, where i m 7 Here, x i,i m in Eqs. 6 and 7 modal value for dummy variables and the mean value for the interval variable. Beyond harm regressions, injury severity is of interest. An ordered probability model is appropriate for modeling the abbreviated injury scale because the injury severity data are ordinal and categorical. The ordered logit model in this study estimated an underlying score as a linear function of the independent variables and a set of thresholds. The probability of observing a certain category of the dependent variable corresponds to the probability that the estimated linear function, plus random error, is within the range of the thresholds estimated for the dependent variable. The equations for ordered logit regression are as follows: k y * = i x i where y * =unobserved and the rest of the terms are defined as above. The observed dependent variable is y, a measure of injury severity, coded as 0, 1, 2, for AIS. It is related to y * as follows: y =0 ify * 0 y = 1 if 0 y * 1 y =2 if 1 y * 2 y = j if j 1 y * 8 9 j=ordinal categories and j =thresholds to be estimated with the explanatory variables. If the error terms is assumed to be normally or logistically distributed, then Pr y =0 = i x i Pr y =1 = 1 i x i i x i Pr y =2 = 2 i x i 1 i x i... Pr y = j =1 j 1 i x i 10 Since the ordered logit model is nonlinear, logit coefficients do not directly measure the marginal effects. Here, marginal effects correspond to the difference in probability when there is a discrete change in the indicator variable from 0 to 1. In order to interpret the results of logit models, the estimated parameters are reported along with marginal effects. Each case in NASS CDS links to a case weight, which can produce national estimates. The weights result from the stages of sample selection and reflect that crash s probability of selection. Thus, our 3-year car SUV crash sample represents the total estimates of car SUV crashes for 3 years instead of the annual estimates. Case weights in the final sample have a range from 2 to 35,894, with a mean of 746 and a standard deviation of 2,448. The median value of the case weights is 172. This wide range of weights leads to the total 402,306 weighted cases based on 539 unweighted car SUV crashes in the database. In addition to weighting cases, the study takes into account the NASS CDS hierarchical stratified sample design to better estimate statistical significance. NASS CDS is a weighted sample; therefore, cases cannot be simply removed. The statistical software package STATA 8.0 allows us to use a subset of an entire sample. It also captures the complex characteristics of NASS CDS sample design using the svyset command. Every year, NASS CDS selects 24 primary sampling units PSUs out of 1195 PSUs total in the United States. Each PSU is a geographic area with minimum population of approximately 50,000 either a central city, an entire county, or a group of contiguous counties. Further, the CDS sampling scheme included stratified sampling strata based on the values of six variables, including vehicle type, KABCO injury scale, disposition of the injured, hospitalization, tow status, and vehicle model year. NASS-CDS identified ten police accident report PAR sampling strata. Definitions for each stratum are given in the Analytical user s manual NASS-CDS The svyset command in STATA takes the sample design into account in calculating confidence intervals and statistical significance. 96 / JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008

6 Table 3. Descriptive Statistics: Independent Variables in Car SUV Analysis Variable and categories Unweighted frequency Weighted frequency Weighted percentage % Mean Standard deviation Min Max Number of passengers , Curb weight difference SUV car , a Driver distraction Car driver SUV driver Nondistracted Nondistracted , Distracted Nondistracted 52 26, Nondistracted Distracted 40 30, Distracted Distracted Missing data Missing data , b Critical precrash events a Car SUV Control loss Car encroaching Outside designated lane 48 17, Turning left , Crossing over 59 56, Turning left Other vehicle in lane 7 17, Other vehicle in lane Other vehicle in lane 52 45, SUV encroaching Control loss Outside designated lane 32 20, Turning left 74 58, Turning right 3 8, Crossing over 51 30, Crossing over Crossing over 17 13, Turning left Crossing over 12 6, c Roadway locations Noninterchange area and nonjunction , Interchange area related 21 23, Intersection related , Driveway, alley access related 48 37, d Roadway types Not divided and Not divided , Not divided and divided without barrier 49 33, Not divided and divided with barrier 7 2, Not divided and one-way traffic 8 1, Divided without barrier and divided without barrier 74 47, Divided without barrier and divided with barrier 6 2, Divided without barrier and one way traffic 10 4, Divided with barrier and divided with barrier 23 12, Divided with barrier and one-way traffic 6 3, One-way traffic and one-way traffic 14 40, e Collision patterns Rear-end 54 50, Head-on 56 20, Angle , Sideswipe, same direction 14 42, Sideswipe, opposite direction 32 14, Note: Total number of unweighted cases is 539, and total number of weighted cases is 402,306. a Critical precrash events: categories whose frequencies are less than 1 are excluded. 98 / JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008

7 Table 4. Comparison of Maximum AIS and Harm by Crash Factors Categories Median max. AIS Mean harm Mean harm per person Sum of harm a Driver distraction a Car driver SUV driver Nondistracted Nondistracted 1 90,890 30,886 5,213,201,577 Distracted Nondistracted 1 149,234 41,831 1,972,570,613 Nondistracted Distracted 1 59,792 24, ,208,460 Distracted Distracted 2 203,267 98,480 63,111,743 Missing data Missing data 1 131,225 47,157 15,090,169,503 b Critical precrash events a Car SUV Control loss Car encroaching 1 180,900 53, ,489,726 Outside designated lane 1 368, ,296 3,203,952,257 Turning left 1 135,093 44,041 3,484,544,605 Crossing over 2 89,815 29,210 2,534,772,417 Turning left Other vehicle in lane 1 19,011 8, ,861,876 Other vehicle in lane Other vehicle in lane 1 72,265 26,532 1,629,463,988 SUV encroaching Control loss 1 48,960 14, ,664,063 Outside designated lane 1 254,328 88,678 2,652,471,877 Turning left 1 56,972 22,027 1,673,953,725 Turning right 1 53,671 9, ,359,465 Crossing over 1 126,532 56,972 1,960,871,841 Crossing over Crossing over 1 186,565 83,539 1,291,515,453 Turning left Crossing over 1 79,061 26, ,355,227 c Roadway locations Noninterchange area and nonjunction 1 181,657 62,670 7,612,977,161 Interchange area related 1 37,865 13, ,370,266 Intersection related 1 105,935 38,703 13,604,512,472 Driveway, alley access related 1 83,902 21,027 1,581,401,997 d Roadway types Not divided and not divided 1 115,895 39,212 14,698,965,208 Not divided and divided without barrier 1 183,079 77,440 3,089,275,277 Not divided and divided with barrier 1 157,744 62, ,740,551 Not divided and one-way traffic 2 147,378 44,283 78,336,773 Divided without barrier and divided without barrier 1 116,911 39,136 2,754,399,161 Divided without barrier and divided with barrier 1 95,769 29, ,606,630 Divided without barrier and one-way traffic 0 70,697 25, ,138,730 Divided with barrier and divided with barrier 1 152,219 50, ,713,308 Divided with barrier and one-way traffic 2 84,416 30, ,359,891 One way traffic and one-way traffic 1 49,986 18,846 1,009,726,367 e Collision patterns Rear-end 1 98,747 38,017 2,475,086,927 Head-on 1 397, ,677 4,050,417,542 Angle 1 99,354 35,316 13,644,877,656 Sideswipe, same direction 1 38,850 12, ,313,738 Sideswipe, opposite direction 1 316,397 97,411 2,248,566,034 a Critical precrash events: categories whose frequencies are less than 1% are excluded. The mean harm of sideswipes in the opposite direction is eight times as large as that of sideswipes in the same direction. Thus, the average harm of control loss by SUVs is much lower than that of control loss by cars. These comparisons provide valuable insights but are not a substitute for regression models, which are presented next. Harm Models To study harm, Table 5 shows the coefficients and marginal effects from two estimated regression models. The ordinary least squares OLS model explains only about 7% of the variation in the data, which is reasonable. The log-transformed regression JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008 / 99

8 Table 5. Linear Regression Model and Log-Transformed Regression Model of Total Harm Model 1 Model 2 a Variable and categories Coefficient Coefficient Marginal effects Number of passengers 24,790 b d 8,259 Curb weight difference SUV car Car driver SUV driver Nondistracted Nondistracted r r Distracted Nondistracted 52, ,433 Nondistracted Distracted 22, Distracted Distracted 43, d 56,336 Missing data Missing data 52, Car event SUV event Control loss Car encroaching 40, Beyond own lane Car encroaching 218, ,799 Turning left Car encroaching 18, ,894 Crossing over Car encroaching 13, c 21,297 Turning left Other vehicle in lane 56, c 24,315 Other vehicle in lane Other vehicle in lane 143, d 25,489 SUV encroaching Control loss 52, ,253 SUV encroaching Beyond own lane 89, ,449 SUV encroaching Turning left 37, b 22,994 SUV encroaching Turning right 112, ,032 SUV encroaching Crossing over 30, b 21,250 Crossing over Crossing over 42, ,195 Turning left Crossing over 64, ,696 a Roadway locations Noninterchange and nonjunction r r Interchange area related 136,295 c ,313 Intersection related 119, ,275 Driveway, alley access related 130, ,653 b Roadway types Not divided and not divided r r Not divided and divided, no barrier 78, b 21,773 Not divided and divided with barrier 41, d 108,003 Not divided and one-way traffic 62, c 128,847 Divided, no barrier and divided, no barrier 17, Divided, no barrier and divided with barrier 54, ,128 Divided, no barrier and one-way traffic 62, ,445 Divided with barrier and divided with barrier 84, ,933 Divided with barrier and one-way traffic 13, ,637 One way traffic and one-way traffic 53, ,263 c Collision patterns Rear-end 69, d 95,306 Head-on 62, b 52,257 Angle r r Sideswipe, same direction 224, ,392 Sideswipe, opposite direction 84, ,720 Constant 110, Number of observations R square/adjusted R square Note: r reference category. a Using In total harm in an accident as dependent variable. Marginal effects are for discrete change of dummy variable from 0 to 1. The coefficients and marginal effects are estimated using weighted data, while t-statistics are estimated using unweighted data. b p 0.1. c p d p / JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008

10 Table 6. Ordered Logit Regression of AIS and Estimated Marginal Effects Marginal effects injury level a Variable and categories Coefficient None Minor Moderate Serious Severe Critical Max Curb weight difference SUV car Car driver SUV driver Nondistracted Nondistracted r Distracted Nondistracted b Nondistracted Distracted b Distracted Distracted b Missing data Missing data Car event SUV event Control loss Car encroaching Beyond own lane Car encroaching Turning left Car encroaching Crossing over Car encroaching c Turning left Other vehicle in lane Other vehicle in lane Other vehicle in lane c SUV encroaching Control loss SUV encroaching Beyond own lane SUV encroaching Turning left SUV encroaching Turning right SUV encroaching Crossing over Crossing over Crossing over Turning left Crossing over a Roadway locations Noninterchange area and nonjunction r Interchange area related Intersection related Driveway, alley access related b Roadway types Not divided and not divided r Not divided and divided, no barrier Not divided and divided with barrier Not divided and one-way traffic Divided, no barrier and divided, no barrier Divided, no barrier and divided with barrier Divided, no barrier and one-way traffic Divided with barrier and divided with barrier Divided with barrier and one-way traffic One way traffic and one-way traffic c Collision patterns Rear-end Head-on Angle r Sideswipe, same direction Sideswipe, opposite direction Number of observations 539 unweighted cases, 402,306 weighted cases Log-likelihood function Pseudo R 2 /Wald /53.16 Note: r reference catagory. a Marginal effects are for discrete change of dummy variable from 0 to 1. The coefficients and marginal effects are estimated using weighted data, while t-statistics are estimated using unweighted data. b p 0.1. c p / JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008

12 Dec. 5, Hollowell, W., and Gabler, H The aggressivity of light trucks and vans in traffic crashes. Paper No , SAE-Society of Automotive Engineers, Troy, Mich. Khattak, A., and Rocha, M Are SUVs supremely unsafe vehicles? Analysis of rollovers and injuries. Transportation Research Record. 1840, Transportation Research Board, Washington, D.C., Kweon, Y.-J., and Kockelman, K Overall injury risk to different drivers: Combining exposure, frequency, and severity models. Accid. Anal Prev., 35 4, McGinnis, R., Davis, M., and Hathaway, E Longitudinal analysis of fatal run-off road crashes, 1975 to Transportation Research Record. 1746, Transportation Research Board, Washington, D.C., National Automotive Sampling System, Crashworthiness Data System NASS-CDS Analytical user s manual, 1997 file, aman1997.pdf May 2, National Center for Statistics and Analysis Motor vehicle traffic crash injury and fatality estimates, 2002 early assessment. DOT HS , pdf Jan. 3, Popkin, C., Campbell, B., Hansen, A., and Stewart, J Analysis of the accuracy of the existing KABCO injury scale. Research Rep., Univ. of North Carolina, Highway Safety Research Center, Chapel Hill, N.C. Summers, S., Prasad, A., and Hollowell, W NHTSA S research program for vehicle aggressivity and fleet compatibility. Paper No. 249, Washington, D.C. Ulfarsson, G., and Mannering, F Differences in male and female injury severities in sport-utility vehicle, minivan, pickup and passenger car accidents. Accid. Anal Prev., 36 2, Univ. of North Carolina Highway Safety Research Center UNC-HSRC North Carolina traffic crash statistics in brief. Chapel Hill, N.C., May 2, Wenzel, T., and Ross, M Are SUVs safer than cars? Analysis of risk by vehicle type and model. Proc., Transportation Research Board 82nd Annual Meeting, TRB, Washington, D.C., Paper No / JOURNAL OF TRANSPORTATION ENGINEERING ASCE / FEBRUARY 2008

### On Road Bicycle Facilities and Cyclist Injury in Bicycle Motor Vehicle Crashes in Chicago CHRISTOPHER MICHAEL QUINN

On Road Bicycle Facilities and Cyclist Injury in Bicycle Motor Vehicle Crashes in Chicago BY CHRISTOPHER MICHAEL QUINN BA, University of Illinois at Chicago, 2008 THESIS Submitted as partial fulfillment

### City of Toronto Bicycle/Motor-Vehicle Collision Study

City of Toronto Bicycle/Motor-Vehicle Collision Study Works and Emergency Services Department Transportation Services Division Transportation Infrastructure Management Section 2003 City of Toronto Bicycle/Motor-Vehicle

### Accident Analysis and Prevention

Accident Analysis and Prevention 42 (2010) 213 224 Contents lists available at ScienceDirect Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap Rainfall effect on single-vehicle

### GAO MOTORCYCLE SAFETY. Increasing Federal Funding Flexibility and Identifying Research Priorities Would Help Support States Safety Efforts

GAO United States Government Accountability Office Report to Congressional Committees November 2012 MOTORCYCLE SAFETY Increasing Federal Funding Flexibility and Identifying Research Priorities Would Help

### How Dangerous Are Drinking Drivers?

How Dangerous Are Drinking Drivers? Steven D. Levitt University of Chicago and American Bar Foundation Jack Porter Harvard University We present a methodology for measuring the risks posed by drinking

### Montana Traffic Safety Problem Identification FFY 2011

Montana Traffic Safety Problem Identification FFY 2011 2009 Data State Highway Traffic Safety Office Montana Department of Transportation 2701 Prospect Avenue Helena, Montana 59620-1001 http://www.mdt.mt.gov/safety/safetyprg.shtml

### Motorcycle Crashes into Road Barriers: the Role of Stability and Different Types of Barriers for Injury Outcome

Motorcycle Crashes into Road Barriers: the Role of Stability and Different Types of Barriers for Injury Outcome Matteo Rizzi, Johan Strandroth, Simon Sternlund, Claes Tingvall, Brian Fildes Abstract This

### Evidence that Seat Belts are as Effective as Child Safety Seats in Preventing Death for Children aged Two and Up *

Evidence that Seat Belts are as Effective as Child Safety Seats in Preventing Death for Children aged Two and Up * Steven D. Levitt University of Chicago Department of Economics, Initiative on Chicago

### Investigation of the Use of Mobile Phones While Driving

Investigation of the Use of Mobile Phones While Driving Prepared by Alasdair Cain Mark Burris Center for Urban Transportation Research College of Engineering, University of South Florida 4202 E. Fowler

### PAY AS YOU DRIVE AUTO INSURANCE

Page 0 PAY AS YOU DRIVE AUTO INSURANCE IN MASSACHUSETTS A RISK ASSESSMENT AND REPORT ON CONSUMER, INDUSTRY AND ENVIRONMENTAL BENEFITS NOVEMBER 2010 AUTHORED BY: MIT Professor Joseph Ferreira, Jr. & Eric

### Motorcycle Helmet Use and Head and Facial Injuries

DOT HS 811 208 October 2009 Motorcycle Helmet Use and Head and Facial Injuries Crash Outcomes in CODES-Linked Data This document is available to the public from the National Technical Information Service,

### DRIVER ATTRIBUTES AND REAR-END CRASH INVOLVEMENT PROPENSITY

U.S. Department of Transportation National Highway Traffic Safety Administration DOT HS 809 540 March 2003 Technical Report DRIVER ATTRIBUTES AND REAR-END CRASH INVOLVEMENT PROPENSITY Published By: National

### THE SAFETY CHALLENGE OF INCREASED CYCLING. Jack Short and Brian Caulfield 1

*Title page (with author's information) Click here to view linked References THE SAFETY CHALLENGE OF INCREASED CYCLING Jack Short and Brian Caulfield 1 Department of Civil, Structural and Environmental

### Traffic fatalities and injuries: the effect of changes in infrastructure and other trends

Accident Analysis and Prevention 35 (2003) 599 611 Traffic fatalities and injuries: the effect of changes in infrastructure and other trends Robert B. Noland Department of Civil and Environmental Engineering,

### A Bayesian Modeling Approach for Cyclist Injury Risk Analysis at Intersections and Corridors

A Bayesian Modeling Approach for Cyclist Injury Risk Analysis at Intersections and Corridors Jillian Strauss PhD Student Department of Civil Engineering and Applied Mechanics McGill University Macdonald

### Evaluating the Relationship Between Near-Crashes and Crashes: Can Near-Crashes Serve as a Surrogate Safety Metric for Crashes?

DOT HS 811 382 October 2010 Evaluating the Relationship Between Near-Crashes and Crashes: Can Near-Crashes Serve as a Surrogate Safety Metric for Crashes? DISCLAIMER This publication is distributed by

### Understanding Bicyclist-Motorist Crashes in Minneapolis, Minnesota

Understanding Bicyclist-Motorist Crashes in Minneapolis, Minnesota A comprehensive look at crash data from 2000-2010 and recommendations for improved bicyclist safety Bicycle and Pedestrian Section Public

### Estimating the Risk of Collisions between Bicycles and Motor Vehicles at Signalized Intersections

Estimating the Risk of Collisions between Bicycles and Motor Vehicles at Signalized Intersections Yinhai Wang (corresponding author) Department of Civil Engineering University of Washington Box 352700,

### The Efficiency of Medical Malpractice Law. Theory and Empirical Evidence. Reed Neil Olsen

The Efficiency of Medical Malpractice Law Theory and Empirical Evidence Reed Neil Olsen Department of Economics Southwest Missouri State University Phone: 417-836-5379 Fax: 417-836-4236 E-Mail: rno174f@smsu.edu

### THE ROAD SAFETY MONITOR 2008 Motorcyclists

T r a f f i c i n j u r y r e s e a r c h F o u n d a t i o n THE ROAD SAFETY MONITOR 2008 Motorcyclists The knowledge source for safe driving The Traffic Injury Research Foundation The mission of the

### Factors on Motorcycle Injuries in Bali

TRANSPORTATION THE INFLUENCE OF YOUNG AND MALE MOTORISTS ACCIDENT Factors on Motorcycle Injuries in Bali D. M. Priyantha WEDAGAMA Lecturer Department of Civil Engineering Faculty of Engineering Udayana

### Cellular Phone Use While Driving: Risks and Benefits

Cellular Phone Use While Driving: Risks and Benefits Karen S. Lissy, M.P.H. Joshua T. Cohen, Ph.D. Mary Y. Park, M.S. John D. Graham, Ph.D. Harvard Center for Risk Analysis Harvard School of Public Health

### What to Expect from California s New Hands-Free Law

Occasional Papers What to Expect from California s New Hands-Free Law Jed Kolko May 2008 The Public Policy Institute of California is dedicated to informing and improving public policy in California through

### Predicting Accident Rates for Cyclists and Pedestrians

Predicting Accident Rates for Cyclists and Pedestrians S. A. Turner, A. P. Roozenburg Beca Infrastructure Ltd, Christchurch, New Zealand T. Francis Francis and Cambridge Ltd, Christchurch, New Zealand

### The Relationship Between Auto Insurance Rate Regulation and Insured Loss Costs: An Empirical Analysis

The Relationship Between Auto Insurance Rate Regulation and Insured Loss Costs: An Empirical Analysis Abstract Laureen Regan Sharon Tennyson Mary Weiss This study points out a potential unintended effect

### Health and Financial Fragility: Evidence from Car Crashes and Consumer Bankruptcy

CHICAGO COASE-SANDOR INSTITUTE FOR LAW AND ECONOMICS WORKING PAPER NO. 655 (2D SERIES) Health and Financial Fragility: Evidence from Car Crashes and Consumer Bankruptcy Edward R. Morrison, Arpit Gupta,

### DOT HS 811 280 June 2010. Motorcycle Crash Causes And Outcomes: Pilot Study

DOT HS 811 280 June 2010 Motorcycle Crash Causes And Outcomes: Pilot Study DISCLAIMER This publication is distributed by the U.S. Department of Transportation, National Highway Traffic Safety Administration,

### Occupation-Specific Human Capital and Local Labor Markets. Jeffrey A. Groen, U.S. Bureau of Labor Statistics

BLS WORKING PAPERS U.S. DEPARTMENT OF LABOR Bureau of Labor Statistics OFFICE OF EMPLOYMENT AND UNEMPLOYMENT STATISTICS Occupation-Specific Human Capital and Local Labor Markets Jeffrey A. Groen, U.S.