Heckle and Chide: Results of a randomized road safety intervention in Kenya

https://doi.org/10.1016/j.jpubeco.2011.06.008Get rights and content

Abstract

We report the results of a randomized field experiment aimed at improving the safety of long-distance mini-busses or matatus in Kenya. Our intervention combines evocative messages aimed at motivating passengers to speak up against bad driving with a lottery that rewards matatu drivers for keeping the stickers in place. Independent insurance claims data were collected for more than 2000 long-distance matatus before and after the intervention. Our results indicate that insurance claims fell by a half to two-thirds, from a baseline annual rate of about 10%, and that claims involving injury or death fell by 60%. While we are unable to identify the mechanism(s) underlying this effect, the intervention is more cost effective in reducing mortality than other documented public health interventions.

Highlights

► RCT of a road safety campaign encouraging bus users to speak up against bad driving. ► ITT estimates of insurance claims reductions of between one half and two thirds. ► Claims involving injury or death fell by 60%.

Introduction

This paper reports the results of a field experiment aimed at improving road safety in the developing world. The specific context is that of long-distance road transportation services in Kenya, where it is popularly believed that otherwise rational young males are transformed, Jekyll-and-Hyde-like, into irrational death-seekers when they occupy the driver's seat of a minibus, or matatu. Our intervention combines the placement inside vehicles of stickers with messages aimed at motivating passengers to speak up against bad driving with a lottery that rewards matatu drivers for keeping the stickers in place. The intervention appears to be extremely cost-effective in terms of reducing mortality, out-performing virtually all other documented public health interventions.

Long distance transportation services in much of the developing world account for a significant share of road traffic injuries and fatalities, which in turn constitute a large and increasing share of both deaths and the disease burden in the developing world. The World Health Organization (2004) reported that 1.2 million people died from road traffic injuries in 2002, 90% in low- and middle-income countries, about the same number as die of malaria. In addition, between 20 and 50 million people are estimated to be injured or disabled each year. Road traffic accidents constitute the largest share, 23%, of deaths due to injury, nearly twice as many as the 14% due to war and violence combined. And they are projected to grow in importance (Lopez et al., 2006), and will be twice as deadly as malaria by 2030 (Mathers and Loncar, 2006).1,2

Many interventions to reduce road accidents have been undertaken in developed economies, including programs to reduce the volume of driving, to improve the safety features of road networks, and to enforce traffic regulations more effectively.3 Publicity campaigns have focused on educating road users, and some have employed shock therapy to get their message across. For example, an advertising campaign in New Zealand aimed at reducing speeding and drunk-driving, and encouraging the use of safety belts, was found to have an impact on road deaths (Guria and Leung, 2004). Fewer studies of interventions in developing countries exist and while the estimated results of these studies are not causal, measured effects are large. The introduction of speed bumps at certain accident hot-spots in Ghana was associated with a 35% reduction in accidents and a 55% reduction in fatalities (Afukaar et al., 2003). Bishai et al. (2008) found that higher intensity police patrols were associated with a 17% reduction in accident rates in Uganda. Perhaps more creatively, in Bogotá, Colombia, mimes were used to ridicule pedestrians and drivers who flaunted traffic rules.4

The intervention we study was simple and cheap: stickers with evocative messages aimed at passengers and encouraging them to “stand up, speak up” were randomly assigned to just over half of 2276 recruited vehicles. High rates of compliance were ensured by running a monthly lottery among drivers of participating treatment matatus, who could win up to 5000 Kenyan Shillings (about $60, or roughly one week's wages) if their vehicle was found to have all stickers intact upon inspection by our field staff. Our main outcome data were collected independently from four insurance companies that together cover more than 90% of sample vehicles, and who were unaware of our intervention at the time it took place. We use insurance claims data for treatment and control vehicles in the two year window bracketing the insertion of the stickers. We identify an impact on driver behavior that is both statistically significant and economically large. Our intent-to-treat estimates indicate that the intervention is associated with a reduction in insurance claims rates of about a half, from a projected counterfactual annual claims rate of about 10%. Our instrumental variables estimate of the average treatment effect on the treated yields an even higher estimate of the impact among compliers. Furthermore, we find that this result is largely due to a reduction in claim events where the driver was at fault. We also document a large reduction in claims involving injury or death. Whether these results are due to the stickers, the lottery (which, as was explained to the drivers, was not meant to reward safe driving explicitly), or some combination of the two, cannot be discerned from the available data.

The rest of the paper is organized as follows. Section 2 describes the intervention, experimental design, data and empirical strategy. We present the results in Section 3, and conclude in Section 4.

Section snippets

Intervention

In a pre-intervention survey of passengers we found that one third of respondents reported having felt that their life was in danger on a recent matatu trip, but that half had never experienced a life-threatening event. Given this heterogeneity in passenger experiences our intervention included a total of five stickers, with both fear stimuli (graphic images of injuries) and simple text messages, as shown in Fig. 1. The stickers, about 11 by 3 in. in size, were placed on the metal panel between

Effects on insurance claims

A visual summary of the results is presented in Fig. 3, in which the trajectories of claims events per 1000 matatus are shown, separately for vehicles assigned to treatment and control, from the first quarter of 2006 to the second quarter of 2009. The horizontal axis in Fig. 3 measures calendar time. Given the considerable lag of 3–6 months in the digital recording of claims, our data for the first two quarters of 2009 are incomplete.9

Conclusions

We have presented evidence that a very cheap intervention can alter the behavior of drivers in the context of long distance minibus transportation services in Kenya. Our intent-to-treat estimates suggest that the intervention reduced the number of incidents leading to an insurance claim by at least half. Although the estimated effect is very large, we argue that it is nonetheless plausible, as the intervention provides passengers and perhaps drivers with timely and salient reminders of the

References (13)

There are more references available in the full text version of this article.

Cited by (34)

  • Simulated patients and their reality: An inquiry into theory and method

    2022, Social Science and Medicine
    Citation Excerpt :

    This has the practical implication that a patient-education campaign on communication in repeat visits could improve quality of care. It parallels the idea in a road safety study, where researchers put posters in vans used for public transport in Kenya, asking travelers to ‘speak up’ if the driver was driving rashly (Habyarimana and Jack, 2011). As the researchers demonstrated, travelers seldom objected without the stimuli, but the new affordance led them to speak up, reducing accidents and fatalities.

  • The sounds of safety silence: Interventions and temporal patterns unmute unique safety voice content in speech

    2021, Safety Science
    Citation Excerpt :

    However, safety silence is different from other behavioural safety (e.g., hand-washing, performing checklists) and voice/silence concepts because it involves the withholding of concerns in any hazardous setting. Moreover, it extends to non-employees (e.g., patients reporting on deteriorating health, minibus passengers speaking up about poor driving; Entwistle et al., 2010; Habyarimana & Jack, 2011). Therefore, safety silence requires a distinct conceptualisation to understand the types of behaviours that constitute the phenomenon and their relationship to safety voice in order to reduce silence during critical incidents.

  • Delivering behavioural change at scale: What conservation can learn from other fields

    2021, Biological Conservation
    Citation Excerpt :

    Well-designed policies can harness social norms to trigger behavioural change at scale and at low cost. For example, in Kenya, messages on stickers authorising bus passengers to reprimand their drivers for reckless driving reduced insurance claims involving injury or death by 50% compared to a randomly allocated control group (Habyarimana and Jack, 2011). Similarly, messages emphasising socially acceptable water use have effectively been used to trigger social norms to reduce household water consumption (Ferraro and Price, 2013; see Abrahamse and Steg, 2013 for a meta-analysis on social influence and resource conservation).

  • Informing and inquiring: Experimental evidence on reducing traffic violations

    2020, Journal of Development Economics
    Citation Excerpt :

    First, it contributes to the literature on reducing traffic violations. Previous studies have explored various strategies for reducing traffic violations, including installing red-light cameras, increasing the severity of punishment, informing drivers of their outstanding traffic tickets, motivating people to speak up about bad driving, or using social comparisons to nudge safe driving (Retting et al., 1999a, 1999b; Bar-Ilan; Sacerdote, 2004; Habyarimana and Jack, 2011; Hansen, 2015; Lu et al., 2016; Chen et al., 2017). This study explores the effects of two information treatments and an inquiry treatment and finds that seeking the public opinion on road safety management can reduce traffic violations.

  • Speaking up to prevent harm: A systematic review of the safety voice literature

    2019, Safety Science
    Citation Excerpt :

    Similarly, indicating the need for a context-agnostic framework, an industry-bias exists towards research in healthcare (n = 41). Further research was conducted in transport (Habyarimana and Jack, 2011; Tucker et al., 2008), aviation (Bienefeld and Grote, 2012), wildland firefighting (Barton and Sutcliffe, 2009), secondary education (Turner et al., 2015), oil and gas (Conchie et al., 2012) or across industries (n = 2; Barnett, 1992; Manapragada and Bruk-lee, 2016 (Study 1)). Furthermore, despite inclusive inclusion criteria for contexts, no studies were found outside of institutions with, arguably, mountaineering groups (Anicich et al., 2015) being the least institutional in nature.

View all citing articles on Scopus

We gratefully acknowledge the financial support of the Center for Global Development and the Safaricom Foundation, and thank Channa Commanday and Bright Oywaya of ASIRT-Kenya, the Kenyan branch of the Association for Safe International Road Travel, an international NGO. We thank Mr. Tom Gichuhi of the Association of Kenyan Insurers, senior executive officers of four large Kenyan insurance companies, and executive officers of the 21 matatu savings and credit cooperatives who assisted us in this project. We also thank the Editor and two anonymous referees, as well as Nada Eissa, David Evans, Luca Flabbi, Garance Genicot, Vijaya Ramachandran, Roger Lagunoff and Tavneet Suri for discussions, and seminar participants at Georgetown, the World Bank, and the Kenya Medical Research Institute. We also acknowledge the pro-bono contributions of George Wanjohi and Saracen Media in Nairobi, and John Wali and volunteers from Junior Achievement Kenya. We thank Lauren Marra for excellent research assistance. Finally we thank Philomena Wanjiru, David Gitahi, Asman Wesonga Suleiman and Nadeem Karmali for their tireless and professional work in leading our team of 20 field workers in implementing the study. All errors are our own.

View full text