Costs of accidents make up an important part of the total external cost of traffic. A substantial proportion of accident costs is related to fatal accidents. In the evaluation of fatal accident costs the availability of an estimate of the economic value of a statistical life is pivotal. We present an overview of the empirical literature on the value of statistical life in road safety (VOSL), and use meta-analysis to determine variables that explain the variation in VOSL estimates reported in the literature. We show that the magnitude of VOSL estimates depends on the value assessment approach (particularly, stated versus revealed preference), and for contingent valuation studies also on the type of payment vehicle and elicitation format. We explain that VOSL estimates cannot simply be averaged over studies. The magnitude of VOSL is intrinsically linked to the initial level of the risk of being caught up in a fatal traffic accident and to the risk decline implied by the research set-up.