Impact evaluation in road safety Patricia Yañez Pagans Inter-American Development Bank Río de Janeiro, Junio 22, 2015
1.3 million people killed annually Up to 50 million injured annually Picture: Michael Gil
Distribution of deaths reported annually in road crashes High income countries 10% Low and middle income countries 90% 3
Road safety investments can have large payoffs Picture: Vicky Thomas Road crashes cost: 65 US$ billion annually Lifting poor above 10 US$/day: 36.74 US$ billion Picture: Douglas Jones Jr.
We need to take action! Picture: Colin Kinner How much do we know about the effectiveness of road safety interventions? What are the limitations (methods and data) and where do we need to build more evidence?
IDB s strategic areas of work in road safety 1) Human factor: Promoting safe behaviors 2) Infrastructure: Incorporating road safety components in transport projects 3) Institutional and normative framework: Road safety laws and national plans 4) Vehicles: Regulatory safety standards and inspections
HUMAN FACTOR: Road safety education could help to reduce accidents Randomized controlled trials (experiments): Road safety education increases awareness and knowledge (Duperrex et al., 2002). Education improves behavior both after training and several months later. Individualized/small group training are more effective than classroom training, virtual or video training (Schwebel et al. 2014) No studies available from developing countries No evidence education actually reduces risk of accidents Few credible randomizations and small sample sizes
HUMAN FACTOR: Awareness campaigns can influence on driver s behavior Randomly assigning stickers with evocative messages encouraging passengers to stand up for bad driving in Kenyan matatus (minibuses) reduces insurance claims by 67% and claims involving injury or death by 60% (Habyarimana and Jack, 2011) Use of administrative data can open opportunities for evaluation in road safety Low cost interventions might have large returns
INFRAESTRUCTURE: Adapting infraestructure can help to reduce accidents but for who? Before-and-after studies with treatment and control groups: Europe and Australia: converting intersections to roundabouts reduces injury accidents between 30% to 50%, and fatal accidents between 50% to 70% (Elvik, 2007) Belgium: increase of 27% in the number of injury crashes with bicyclists after constructing roundabouts (Daniels et al., 2008) => Different impacts on different people, trade-offs => Small samples, no balance tests => Need for detailed data (tipology and location)
INFRAESTRUCTURE: Technology can help to reduce speed and accidents Before-and-after studies with treatment and control groups: In United Kingdom speed cameras reduce injury accidents by 50%, mainly for pedestrians (Christie et al., 2003). In Pennsylvania, US, lengthening yellow signal timing reduced red light violations by 36%. Adding a red light camera further reduced violations by 96% (Retting et al., 2007) In Virginia, US, red-light cameras decreased red-light running crashes by 42%, but increased rear-end crashes by 27%. Overall, reduction in crash cost (Garber et al. 2007) Combination of interventions, trade-offs
INSTITUTIONAL AND NORMATIVE FRAMEWORK: Laws against drinking and violations can help to reduce accidents Difference-in-Differences approach: In European countries, reducing the illegal blood alcohol content (BAC) level to 0.5 reduces accidents between 6-12% but only for young males living in urban areas (Albalate, 2008). In Chile changes in BAC to 0.3 plus other financial and license revocation penalties reduces the number of alcohol-related car accidents by 18% (Otero, 2013).
VEHICLES: Car inspections may not be enough to prevent accidents In Norway no differences in accident rates were found for cars randomly assigned to inspections, although their technical condition improved compared to those not inspected (Fosser, 1992) => Elements of behavior beyond car technical conditions matter
Methods can be improved Mostly before-andafter approaches Small sample sizes Almost no balance tests Few credible randomizations
Data needs and opportunities Picture: Marino Gonzalez Use of administrative data open opportunities for evaluation (insurance data, crash reports, etc.) Generate or improve data accesibility in developing countries (digitized and easily accesible) Other data will need to be collected through household or individual level surveys (behavior, attitudes).
Key messages Evidence needed for developing countries! Generate or improve data available Key aspects moving forward: Better understanding of behavior Heterogeneous impact effects and mechanisms Identify trade-offs
THANK YOU