Can a road safety measure be both effective and ineffective at the same time?

Transcription

1 Can a road safety measure be both effective and ineffective at the same time? A game-theoretic model of the effects of daytime running lights Rune Elvik, Institute of Transport Economics (re@toi.no)

2 Effects of daytime running lights Several reviews and meta-analyses have been reported: Helmers 1988 Koornstra 1993 Elvik 1996 Koornstra, Bijleveld, Hagenzieker 1997 Elvik, Christensen, Fjeld Olsen 2003 All of them have been criticised there is still no consensus among researchers about the effects of DRL Are there reasons for this doubt? Yes, there are Page 2

3 Anomalies in DRL-studies Intrinsic (individual) effects have become smaller over time: Early studies: cars with DRL are involved in % fewer accidents than cars without DRL New studies: The difference is around 5 % There is no dose-deponse relationship associated with increased use of DRL: Increasing DRL-use from 30 to 90 % does not have a greater effect on accidents than increasing it from 80 to 90 % Some studies of DRL-laws found no effect on accidents: No effect in Denmark and Norway Effects in other countries Page 3

4 Can the anomalies be resolved? Or to put the question in different terms: Can a road safety measure be both effective and ineffective at the same time? Yes, it can A game-theoretic model based on Schelling-diagrams can show how this is possible This offers a new perspective on the DRL-literature It must be stressed, however, that this is not the only possible interpretation of the DRL-literature Page 4

5 Thomas C. Schelling (Wikipedia) Thomas Crombie Schelling (born 14 April 1921) is an American economist and professor of foreign affairs, national security, nuclear strategy, and arms control at the School of Public Policy at University of Maryland, College Park. He is also co-faculty at the New England Complex Systems Institute. He was awarded the 2005 Nobel Memorial Prize in Economic Sciences (shared with Robert Aumann) for "having enhanced our understanding of conflict and cooperation through game-theory analysis". Page 5

6 250 Schelling diagram of benefits of vaccination Benefits (arbitrary units) Vaccinate Totalexternal benefits as a function of the percentage who vaccinate Not vaccinate -100 Percent of population choosing to vaccinate (upper curve) Page 6

7 How can DRL-effects be modelled? Use of daytime running lights is a binary choice (yes or no) Use of daytime running lights has positive internal effects (i.e. it benefits those who use DRL) Use of daytime running lights has negative external effects (i.e. it brings a disadvantage to those not using DRL) If internal and external effects exactly balance, it follows that: Cars using DRL will always be safer than cars not using DRL Changes in the share of cars using DRL will have no effect on accidents Page 7

8 Percentage change in reaction time associated with change in DRL-use Percentage change in reaction time associated with expectations about DRLuse. Derived from Hole and Tyrrell (1995) Change in reaction time for motorcycle with headlights off following 15 pictures of motorcycle with headlights on and 10 pictures of motorcycle with headlights off Change in reaction time for motorcycle with headlights on following 23 pictures of motorcycle with headlights off Change in reaction time for motorcycle with headlights off following 24 pictures of motorcycle with headlights on and 1 picture of motorcycle with headlights off Change in reaction time for motorcyclewith headlights on following 23 pictures of motorcycle with headlights on Percent of pictures showing motorcycle with daytime running lights on Page 8

9 2.500 Relative accident rates of cars with DRL, cars without DRL and all cars Relative accident rate Cars without DRL (upper curve) Cars with DRL (lower curve) Share of cars using DRL Page 9

10 50 Percentage reduction of accidents for cars using DRL compared to cars not using DRL Percentage accident reduction y = x x R² = Publication year of study Page 10

11 30 Relationship between percentage points increase in DRL-use and percent change in the number of accidents 25 Percent change in the number of accidents Percentage points increase in use of daytime running lights Page 11

12 Effects on accidents of DRL-laws by study quality Percentage change in the number of accidents Study quality Page 12

13 There is support for the game-theoretic model Intrinsic effects of DRL are smaller when use of DRL is more widespread There is no dose-response relationship between the percentage using DRL and the size of the effect on accidents The effects of DRL-laws could be zero although findings are inconsistent as far as DRL-laws are concerned The traditional interpretation of the effects of DRL can still be defended Anomalies are not all that uncommon Page 13

14 Concluding reflections A strong case can be made for theoretical modelling of the effects of road safety measures Game-theoretic models can be applied to, for example: Police enforcement (Bjørnskau and Elvik 1992) Daytime running lights Studded tyres The value of theoretic models is to try to make sense of findings that may at first seem counterintuitive Such models help us look for systematic patterns in effects But we will never find a theory of everything Page 14