AbstractProject risk is an important part of managing large projects of any sort. This study contributes to the state of knowledge in project risk management by introducing a data-driven approach to measure risk identification performance using historical data. In the early phases of a project, the identification and assessment of risk is based largely on experience and expert judgment. As a project moves through its life cycle, these identified risks and the assessment of them evolve. Some risks become issues, some are mitigated, and some are retired as no longer important. This study investigated the quality of early risk registers and risk assessments on large transportation projects and compared them to how the identified risks evolved on historical projects. The investigation involved the use of textual analysis of archival risk register documents. Finite-state automation methods akin to Markov chain models were used to track the changes in risk attributes on large infrastructure projects as the projects matured. The objective was to be better able to anticipate how project risks will change as projects move forward and to be better able to forecast changes to the risk register from ex ante to ex post conditions. Results from 11 major US transportation projects suggested that, on average, fewer than 65% of ex ante identified risks ultimately occurred in projects and were mitigated, while more than 35% did not occur and were retired. In addition, more than half of the risks emerged during project execution when new information became available. Based on the categorization of risk management styles, we find that identifying risks early in the project life cycle is necessary, but not sufficient to ensure successful project delivery. A project team with positive doer behavior (i.e., actively monitoring and identifying risks during project execution) performed better in delivering projects on time and within budget.