In this study,loss estimation models were developed for reasonably accurate assessment of economic and human losses from seismic events in the Mediterranean region,based on damage assessment at an urban scale.Data were compiled from existing worldwide databases,and completed with earthquake information from regional studies.Economic data were converted to a single common currency unit (2015 USD value) and the wealth of the areas affected by 65 earthquakes of the region from 1900 to 2015 was assessed.Reduced-form models were used to determine economic and human losses,with earthquake magnitude and intensity as hazard-related variables,and gross domestic product of the affected area and the affected population as exposure-related variables.Damage to buildings was also used as a hazard-related variable to predict economic and human losses.Finally,site-specific regression models were proposed for economic and human losses due to earthquakes in the Mediterranean region,and more specifically,in Algeria.We show that by introducing the damage variable into the models,prediction error can be reduced,and that accuracy of loss model estimation is site dependent and requires regional data on earthquake losses to improve.A case study for Constantine,Algeria shows the improvements needed for increased accuracy.