Traceability systems are information tools implemented within and between firms in food chains to improve logistics and transparency or to reduce total food safety damage costs. Information about location and condition of products is critical when food safety incidents arise. This paper uses a principal-agent model to investigate the optimal choice of voluntary traceability in terms of precision of information on a given attribute at each link of a food chain. The results suggest that four scenarios may emerge for the supply chain depending on the costs of a system and whether or not the industry can internalize total food safety damages : no traceability, traceability for one link, equal traceability for all links, or different positive traceability levels across all links.