This paper uses the experimental setup of Tanaka et al. (2010) to measure refugees’ risk preferences. A sample of 206 asylum seekers was interviewed in 2017-18 in Luxembourg. Contrary to studies which focus on risk aversion in general, we analyze its components using a cumulative prospect theory (CPT) frame-work. We show that refugees exhibit particularly low levels of risk aversion compared to other populations and that CPT provides a better fit for modelling risk attitudes. Moreover, we include randomised temporary treatments provoking emotions and find a small significant impact on probability distortion. Robustness of the Tanaka et al. (2010) experimental framework is confirmed by including treatments regarding the embedding effect. Finally, we propose a theoretical model of refugee migration that integrates the insights from our experimental outcomes regarding the functional form of refugees’ decision under risk and the estimated parameter values. The model is then simulated using the data from our study.