Lec 11, Ch.8: Accident Studies (objectives)

Be able to explain different approaches to traffic safety

Be familiar with typical data items that are collected and stored (through reading)

Know types of accident analysis typically conducted

Know how to conduct site analyses Be familiar with different safety

countermeasures and their cost effectiveness (through reading)

What we cover in class today… Approaches to highway safety Accident data collection and

record systems (quick review) Accident statistics Analysis of site-specific

accident data

Approaches to highway safetyReducing accident occurrence

Reduce driver error by design with good SD, proper signing and marking (traffic control devices) Reducing accident potential

Reducing the severity of accidents

Give drivers time and space to “recover” from errors and to minimize the severity when it occurs. Proper use of guardrail, barriers, impact attenuators, breakaway signposts, etc.

Improving crash survivability

Involves vehicle design, including shock-absorbing bumpers, seat belts, air bags, padded dashboards, etc. You need legislative actions – like mandatory seat belt installation, air bag installation, etc.

Programmatic safety efforts

Federal and state programs that address traffic safety on a policy level, like state vehicle inspection programs, speed limits, 21 year-old drinking age, state DWI programs, etc.

Design aspects of safety

Embed safety in facility design: alignment, roadside, median barriers, gore areas, etc.

Accident data collection and record systems (discussed in the inventory discussion in ch4)

One of the most basic functions of traffic engineering is keeping track of the physical inventory.

Example: AIMS (Accident Info Mgmt System) by JMW EngineeringAccident spot map

Collision diagram

Accident statistics

Occurrence

Involvement Severity

Types of accidents

Numbers of accidents

Categories of vehicles

Categories of drivers

No. of deaths

No. of injuriesTypes of statistics

For accident stats, visit www.nhtsa.dot.gov

Typical accident rates used“Bases” are needed to compare the occurrence of accidents at different sites.

Population based:

Area population (25 deaths per 100,000 pop)

No. of registered vehicles (7.5 deaths per 10,000 registered vehicles)

No. of licensed drivers (5.0 deaths per 10,000 licensed drivers)

Highway mileage (5.0 deaths per 1,000 miles)

Exposure based:

VMT (5.0 deaths per 100 million VMT)

VHT (5.0 deaths per 100 million VHT)

Severity index:

No. of deaths/accident (0.0285 accidents per accident)

No. of injuries/accident

Typical basic accident rates:

general accident rates describing total accident occurrence

fatality rates describing accident severity

involvement rates describing the types of vehicles and drivers involved in accidents

Types of statistical displaysThe purpose of the display dictates the type of display – temporal, spatial, accident type, etc.

Site analysis

The first thing you do is visit the site and prepare a condition diagram of the site.

Purposes:

Identify contributing causes

Develop site specific improvements

Two types of info:

Accident data

Environment & physical condition data

(Table 8-4 is useful.)

Site analysis (cont)Then we prepare a collision diagram.

Site analysis (cont)Group accidents by type and answer the following 3 questions, which will lead you to possible countermeasures.

What drivers actions lead to the occurrence of such an accident?

What conditions existing at the location could contribute toward drivers taking such actions

What changes can be made to reduce the chance of such actions occurring in the future?

Rear-end collisions:

Driver: Sudden stop & Tailgating

Environment: Too many accesses and interactions with vehicles in/out of the accesses, bad sight distance, short/long yellow interval, inappropriate location of stop lines, etc.

Determining high-accident locationsH0: Accident rate at the location under consideration in the group is equal to the average rate of the group.

H1: Accident rate at the location under consideration in the group is higher than the average rate of the group.

This is a one-tailed test. Why?

z = 1.645

5%

s

xxz

1 xsx 645.11

Locations with a higher accident rate than this value would normally be selected for specific study.

Example:

Highway Section 33 has 210 accident/100MVMT. The mean accident rate for the similar classification group = 89 accident/100MVMT, SD = 64 accident/100MVMT. Should an analyst flag Section 33 as hazardous? With 95% confidence level?

Determining high-accident locations: Expected value analysis (from Garber & Hoel)

H0: Accident rate at the location under consideration in the group is equal to the average rate of the group.

H1: Accident rate at the location under consideration in the group is not equal to the average rate of the group (In another words, we are trying to find whether the site under study is “unusual” or not. We are not specifically proving it is “over-represented” or not.)

ZSxEV

Locations with a higher accident rate than this value would normally be selected for specific study.

Note this method is used only to compare sites with similar characteristics.

z = 1.96 for the 95% confidence level

“Over-represented”“Under-represented”

Not over-represented or under-represented

%5.2 %5.2

zsx zsx x

Example 5-3 (modified): An intersection with 14 rear-end, 10 LT, and 2 right-angle collisions for 3 consecutive years

Check about rear-end collisions

34.1046.496.15.140.705.0 toEV Rear-end collisions are over-represented at the study site at 95% confidence level, since 14 > 10.34.

Check about LT collisions

92.1288.096.107.390.605.0 toEV

LT collisions are not over-represented or under-represented at the study site at 95% confidence level, since 0.88<10 < 12.92.

Control site Rear-end

LT collisions

Right-angle

1 8 11 4

2 5 12 5

3 7 4 3

4 8 5 6

5 6 8 7

6 8 3 8

7 9 4 4

8 10 9 5

9 6 7 6

10 7 6 7

Mean 7.40 6.90 5.5

SD 1.5 3.07 1.58

Check about right-angle collisions 65.104.296.158.15.505.0 toEV Right-angle collisions are under-represented at the study site at 95% confidence level, since 2 < 2.4.

Statistical analysis of before-after accident data

BA

BA

ff

ffz

1

Method 1: Use the Normal Approximation method:

z1 = test statistic, 1.96 at the 95% confidence level for a “change”, 1.645 for a “reduction.”

fA = No. of accidents in the “after” study

fB = No. of accidents in the “before” study

This method is however not recommended by the current Manual of Transportation Engineering Studies.

Statistical analysis of before-after accident data (cont)Method 2: The Modified Binomial Test

Example: Before 14 conflicts were observed at a stop-sign controlled intersection. After the installation of a signal, they observed 7 conflicts. Were the signal effective?

Solution: Figure on the right shows that for 14 before conflicts you need a 60% reduction to be significant at the 95% CL. 7/14=50% reduction. So, you cannot reject the null hypothesis.