This paper provides an international overview of the most recent estimates of the social costs of road crashes: total costs, value per casualty and breakdown in cost components. It answers the question to what extent the total social costs of road crashes differ between countries and what the explanations for these differences are. The analysis is based on national reports about costs of road crashes of 16 countries, of which nine high income countries (HICs) and seven low and middle income countries (LMICs). The analysis shows that the share of social costs of road crashes in HICs ranges from 1.0 to 4.6% of the gross domestic product (GDP) with an average of 2.6%. The main explanations for differences between these countries are differences in methodologies regarding valuation of human costs and correction for underreporting. Excluding countries that do not use a ‘willingness to pay’ method (that is internationally recommended) for estimating human costs and countries that do not correct for underreporting, results in an average share of the costs of 3.7% of GDP. For LMICs that do correct for underreporting the share in GDP ranges from 1.2 to 3.0%. However, none of the LMICs included has performed a willingness to pay study into the human costs. A major part of the costs is related to injuries: an average share of 50% for both HICs and LMICs. The share of fatalities in the costs is 20% and 30% respectively. Prevention of injuries is thus important to bring down the socio-economic burden of road crashes. The paper shows that there are a number of methodological differences between countries regarding cost components that are taken into account and regarding the methods used to estimate specific cost components. In order to be able to make sound comparisons of the costs of road crashes across countries, (further) harmonization of cost studies is recommended. This can be reached by updating and improving international guidelines and applying them in future cost studies. The information regarding some cost components, particularly human costs and property damage, is poor and more research into these cost components is recommended.