A new method for improving relative locations of clustered earthquakes is presented and applied to a suite of microearthquakes induced by hydraulic fracturing. The method is based on the assumption that clustering of earthquake hypocenters is obscured by the uncorrelated scatter of individual hypocenters. The method is implemented as an additional constraint in a Joint Hypocenter Determination (JHD) scheme. The method shifts event hypocenters toward the center of mass of the events within some volume surrounding the event location if the RMS misfit between predicted and measured arrival times does not increase significantly. The method uses the same basic assumption of Jones and Stewart (1997), which is that there is greater clustering in actual earthquake locations than there is in locations determined using conventional techniques. Our method differs in that it is implemented as part of the JHD process so it operates on raw travel-time data rather than on derived hypocenters. The method produces hypocenters from a demonstration field dataset that are similar to those obtained by Phillips et al. (1997), from time-consuming precise manual repicking of relative arrival times of events. The clustering constraint can easily be incorporated as an additional constraint in earthquake location/velocity tomography codes and may lead to improved velocity structure determination and earthquake location pattern identification and interpretation.