Pedestrian safety is a growing problem in the USA. The proportion of pedestrian fatalities has steadily increased over the last 10 years from 11% of all traffic fatalities in 2006 to 15% in 2005. This study examined crashes and near‐crashes from the Second Strategic Highway Research Program (SHRP‐2) naturalistic driving study, which is a database containing time series vehicle data and forward‐facing video of real‐world driving interactions. Of the 162 vehicle‐pedestrian events examined, scenarios in which the pedestrian was travelling perpendicular to the vehicle accounted for over 60% of the events. In 31% of the events, some sort of visual obstruction was involved, yet the average duration for which the pedestrian was visible was over two seconds. Drivers initiated braking evasive manoeuvres on average 2.3 seconds prior to the proximity impact point. In about 40% of the events, drivers did not initiate braking until one second before the proximity impact point, regardless of lighting conditions. Based on the assumption that one second is needed to detect and respond to an event, an automatic emergency braking (AEB) system would have sufficient time to avoid or mitigate approximately 90% of the events examined. One second is commonly assumed to be the threshold for AEB systems as the system must detect the pedestrian, alert the driver, and initiate braking protocols, but a range of thresholds were examined. This study presents methods for an examination of characteristics associated with real‐world vehicle‐pedestrian crashes and near crashes to estimate the benefit of pedestrian AEB system in the USA.