Sensitivity of Building Loss Estimates to Major Uncertain Variables Keith A. Porter, a) M.EERI, James L. Beck, b) M.EERI, and Rustem V. Shaikhutdinov c) This paper examines the question of which sources of uncertainty most strongly affect the repair cost of a building in a future earthquake. Uncertain- ties examined here include spectral acceleration, ground-motion details, mass, damping, structural force-deformation behavior, building-component fragility, contractor costs, and the contractor’s overhead and profit.We mea- sure the variation (or swing) of the repair cost when each basic input variable except one is taken at its median value, and the remaining variable is taken at its 10 th and at its 90 th percentile. We perform this study using a 1960s high- rise nonductile reinforced-concrete moment-frame building. Repair costs are estimated using the assembly-based vulnerability (ABV) method. We find that the top three contributors to uncertainty are assembly capacity (the struc- tural response at which a component exceeds some damage state), shaking intensity (measured here in terms of damped elastic spectral acceleration, S a ), and details of the ground motion with a given S a . [DOI: 10.1193/1.1516201] INTRODUCTION Uncertainty is generally costly in earthquake engineering. If one wants to ensure a minimum level of performance with a certain probability, then greater uncertainty in ei- ther seismic demand or capacity increases the level of nominal capacity that must be designed into the system. If one can reduce uncertainty, one can generally reduce cost. But there are many sources of uncertainty in earthquake engineering performance: shak- ing intensity, details of future ground motions with that intensity level, a variety of physical characteristics of the structure in question, construction and repair costs, and details of occupancy and use. One can assess and possibly reduce uncertainties in each one, but which are the im- portant ones deserving the most attention? By important variables, we mean the ones whose uncertainty contributes most strongly to overall uncertainty in seismic perfor- mance. There are two benefits to knowing the relative contribution of each variable to overall uncertainty in a performance metric. First, the variables that do not contribute much to overall uncertainty can reasonably be taken at their best-estimate value, rather a) G. W. Housner Senior Research Fellow, California Institute ofTechnology, Pasadena, CA91125-4400 b) Professor of Civil Engineering and Applied Mechanics, California Institute of Technology, Pasadena, CA 91125 c) Doctoral Candidate in Applied Mechanics, California Institute of Technology, Pasadena, CA 91125 719 Earthquake Spectra, Volume 18, No. 4, pages 719–743, November 2002; © 2002, Earthquake Engineering Research Institute