This paper investigates epistemic uncertainty in the results of seismic hazard analyses for the San Francisco Bay Area and their role in the broader picture of seismic performance assessment. Using the 2002 Working Group on California Earthquake Probabilities earthquake rupture forecast, epistemic uncertainty in the seismic hazard for several different intensity measures and sites in the San Francisco Bay Area is investigated. Normalization of the epistemic uncertainty for various sites and intensity measures illustrates that the uncertainty magnitude can be approximately estimated as a function of the mean exceedance probability. The distribution of the epistemic uncertainty is found to be dependent on the set of alternative ground-motion prediction equations used but is frequently well approximated by the lognormal distribution. The correlation in the hazard uncertainty is observed to be a function of the separation between the two different intensity levels, and a simple predictive equation is proposed based on the data analyzed. Three methods for the propagation of seismic hazard epistemic uncertainty are compared and contrasted using an example of the 30-year collapse probability of a structure. It is observed that, for this example, epistemic uncertainty in the collapse capacity is more influential than that in the seismic hazard