This paper formulates a procedure for estimating the consequence of an earthquake prediction on statistical estimates of seismic risk. Several examples, using this procedure, indicate the following general conclusions: if a prediction with a confidence level of 1.0 significantly affects the risk at a site, then the same prediction at smaller confidence levels will also affect the risk. If an earthquake is predicted in a high risk zone, it can be important for an entire adjacent low risk zone; a prediction for the adjacent low risk zone may not affect the high risk zone at all. Finally, a prediction for a region which may significantly affect the amplitudes of shaking that can be expected in the next year may have no importance for structures which are designed (or planned) with a 50 or more year lifetime, depending on the size of the predicted earthquakes.
These conclusions are drawn from a hypothetical, yet plausible, case study in the Los Angeles, California, region, where hypothetical predictions are considered for events on the San Andreas fault, the Sierra Madre fault, and a region of about 4300 km2 covering the most populated region of the Los Angeles basin. For the background seismicity model used in this study, the threshold of a significant (to the seismic risk) prediction for 1-yr risk is M = 5 on the San Andreas fault, M = on the Sierra Madre fault, and M = 4 in the Los Angeles zone. For the 50-yr risk, the thresholds of significance are M = 7, M = , and M = 6, respectively, for the same three zones.