Table 2.
Indicator | Mean | SD | Minimum | Maximum |
Claims data (n = 9,358 incidents)* | ||||
No. of claims per incident | 1.993 | 3.516 | 1 | 83 |
No. of fatalities per incident | 0.105 | 0.49 | 0 | 21 |
Share of incidents with at least one fatality | 0.078 | 0.269 | 0 | 1 |
No. of claims with broken bones per incident | 0.155 | 0.517 | 0 | 11 |
Share of incidents with at least one injury | 0.444 | 0.497 | 0 | 1 |
No. of claims with soft tissue injury per incident | 1.051 | 3.114 | 0 | 77 |
Share of incidents with nonvehicle claimants | 0.188 | 0.39 | 0 | 1 |
Share of incidents with vehicle damage claimant | 0.256 | 0.436 | 0 | 1 |
Passenger survey (n = 9,807)† | ||||
Share male | 0.68 | 0.46 | 0 | 1 |
Age | 32.69 | 8.38 | 13 | 83 |
Share speak Kikuyu | 0.35 | 0.48 | 0 | 1 |
Share speak Meru | 0.25 | 0.43 | 0 | 1 |
Share had at least two bus trips in past week | 0.58 | 0.49 | 0 | 1 |
Share ever exposed to radio spot | 0.65 | 0.48 | 0 | 1 |
Trips data‡ | ||||
Speed (n = 4,405) | ||||
Distance covered on trip, km | 69.07 | 50.97 | 3 | 229 |
Maximum speed reached, km/h | 95.95 | 15.51 | 25 | 155 |
Moving average speed, km/h | 45.88 | 17.04 | 3 | 105 |
Events (n = 1,471) | ||||
Share of events reported as reckless | 0.15 | 0.36 | 0 | 1 |
Share of events passengers speak to driver | 0.18 | 0.38 | 0 | 1 |
Claims data for January 1, 2009–December 13, 2013, during which period there were 9,538 incidents. An incident is a crash or other event that results in at least one insurance claim. There can be multiple claims (e.g., by different injured individuals) associated with a single incident. Data reflect all claims, not just claims by vehicles in our experimental sample.
Kikuyu and Meru are dominant languages spoken in Central Kenya, where the two radio stations that aired the safety messages have the largest audiences.
Enumerators collected speed data automatically on GPS units; 4,405 valid records were submitted. However, events such as excessive speed, reckless driving, and passenger complaints to drivers were recorded manually. Of these, 1,471 events could be properly matched to vehicles.