The objective of this study consists in analyzing the determinants of the internal mobility of refugees in Turkey. We track down this mobility relying on geolocalized mobile phone calls data and bring these measures to a micro-founded gravity model in order to estimate the main drivers of refugee mobility across 26 regions in 2017. Our results show that the movements of refugees are sensitive to income differentials and contribute therefore to a more efficient allocation of labor across space. Comparing these findings with those of individuals with a non-refugee status, we find that refugees are more sensitive to variations of income at origin and to distance, while less responsive to changes in income at destination. These findings are robust to the way mobility is inferred from phone data and to the choice of the geographical unit of investigation. Further, we provide evidence against some alternative explanations of mobility such as the propensity to leave refugee camps, transit through Turkey, social magnet effects and sensitivity to agricultural business cycles.