How accurately does the public perceive differences in

transport risks?

TRB-paper 06-0148Rune Elvik and Torkel Bjørnskau

Published in Accident Analysis and Prevention 2005, 1005-1011

2

Survey of perceived risk of travel in Norway

• First made in 2000• Repeated in 2003• Sample size was 1,000• Basic question:

– How safe do you think it is to travel by (car, bus, plane…)?

– Very safe, safe, a little unsafe, very unsafe

3

Transport modes covered

• Car (role not specified)• Car driver• Motorcycle• Bus• Walking• Cycling• Commercial aviation• Commercial ship (not recreational boats)

4

What do we know?

• Risk perception in the small and in the large

• The small is about specific hazards encountered when, e.g., driving

• The large is about societal risks, faced by everybody (not personal risks)

• Knowledge regarding relative risks is fairly accurate (modes are ranked correctly)

5

What do we know, continued

• The absolute number of accident victims is less accurately known

• Risks stated as fatality rates (5 x 10-5) are poorly understood

• Differences in risks tend to be underestimated

6

An initial observation

• The distribution of answers between the categories (very safe, safe, a little unsafe, very unsafe) seemed to agree pretty well with actual differences in risk

• Actual differences = differences between modes in terms of fatality rate per billion kilometres of travel

7

Conversion of answers to scales representing

perceived risk• Can be done in many ways• Four scales were developed:

– A 100 (very unsafe) to 1 (very safe) scale– A 4 (very unsafe) to 1 (very safe) scale– A 10 (very unsafe) to 0.01 (very safe) scale– Odds of answering unsafe/safe

• The relationship between the mean values of these scales and fatality rate or injury rate was studied

8

Comparison of the range of scales

0.001

0.01

0.1

1

10

100

1000

Actual risk100 to 1

scale 4 to 1 scale 10 to 0.01 scale

Odds scale

9

Comparison of actual and perceived risk

0

20

40

60

80

100

120

Aviation Railway Ship Bus Car Motorcycle Cycle Walking

Re

lati

ve

fa

tali

ty r

ate

(s

afe

st m

od

e =

1)

0

5

10

15

20

25

30

Rel

ativ

e p

erc

eive

d f

ata

lity

ra

te a

cc

ord

ing

to

0.0

1 to

10

sc

ale

(s

afe

st m

od

e =

1)

Actual risk

Perceived risk

10

Actual and perceived car occupant injury rates

0

1

2

3

4

5

6

7

8

18-24 25-34 35-44 45-54 55-64 65-74 75 +

Rel

ativ

e in

jury

ra

te (

sa

fes

t =

1)

0

1

2

3

4

5

6

7

8

Rel

ativ

e p

erc

eive

d i

nju

ry r

ate

ac

co

rdin

g t

o 0

.01

to

10

sca

le (

safe

st =

1)

Actual - men

Actual - women

Perceived - men

Perceived - women

11

Correlations between actual and perceived risk

Correlations between statistical risk and perceived risk

Way of representing answers to questions about perceived risk

Comparison with respect to

Scale, range 1 to 100

Scale, range 1 to 4

Scale, range 0.01 to 10

Odds for unsafe to safe

Mode of transport .858 .887 .817 .749

Car occupants, men .450 .068 .771 .096

Car occupants, women -.453 -.422 .776 -.739

Car driver, men .426 -.034 .694 .160

Car driver, women -.666 -.535 .361 -.724

Cycling, men .474 .289 .370 .622

Cycling, women -.396 -.435 -.135 -.202

Walking, men .382 -.160 .518 .424

Walking, women .924 .494 .492 .981

Mean correlation .222 .016 .518 .151

12

Main conclusions

• The Norwegian public has an accurate perception of differences in risk between transport modes

• The range of values on scales developed to represent perceived numerically tends to be narrower than the range of actual differences in risk

13

Conclusions, continued

• The 10 to 0.01 scale correlates best with differences in statistically estimated risk

• Optimism bias is found: motorcycle riders think motorcycling is safer than the general public

• Recent major accidents appear to influence risk perception