Traffic microsimulation is a commonly used tool in traffic engineering. Given its flexibility and cost-efficiency, it is increasingly used for evaluating traffic safety. In real life traffic, unsafety is in many cases due to human error in driving behaviour. In traffic microsimulations however, driving behaviour is highly dependent on the formulated driving behavioural models and the level of their realism. Most of these behavioural models were developed ignoring the inconsistencies and error proneness of human behaviour. A quantitative evaluation of the differences in the safety level between real life and simulated traffic, considering the mathematical formulation of different driver behavioural models, is lacking in the literature. The main aim of this study is to investigate the influence of different behavioural models’ formulations on the correlation between simulated and empirical surrogate safety measures’ outcomes. For this purpose, high-quality empirical trajectory data were used to calibrate and validate different driver behavioural models. SUMO (Simulation of Urban MObility), an open-source traffic microsimulation software, was used as a platform for calibrating, validating, and testing four distinct combinations of car following and lane changing models. The results show that, regardless of the behavioural model formulations used, the number of simulated traffic conflicts is overestimated. This is most likely due to a higher frequency of lane changes and an unrealistic distribution of traffic over the different lanes in microscopic traffic simulation. The severity of the simulated conflicts was shown to be reasonably accurate at an aggregate level but not significantly comparable at a microscopic level. @en