Quantitative biological assessment indices overcome many of the challenges faced when trying to convey the status or trends of complex biological communities with large natural variability, particularly when attempting to evaluate the impacts from human influences. In this paper, we developed a biological condition index for shallow (<30 m) rocky reefs of the Southern California Bight, evaluated its ability to distinguish healthy from stressed sites, and then applied the index by examining relative correlations with fishing and water quality as ecosystem stressors. We utilized a multivariate, predictive index based on the ratio of observed‐to‐expected taxa (O/E). O/E indices are relatively common in freshwater environments, but rarely utilized in marine systems and never before applied to rocky reefs. Based on expectations drawn from region‐wide reference reefs with the least fishing or water quality stress, the O/E index predicts expected taxa at a new site based on environmental factors such as sea surface temperature, reef area, and slope, among others. The observed taxa at that site are then compared to the predicted taxa to generate index scores; values near unity indicate intact, reference‐like communities. Overall, the accuracy of the index was high, with minimal bias, and precision exceeded the performance of an index based on null models (i.e., indices that did not account for natural gradients). The mean index score was significantly higher among reference sites than stressed sites; however, sensitivity was low, as 84% of stressed sites had scores within the range of reference sites. Ultimately, fishing pressure was more correlated with changes in index scores from the non‐reference data set than was water quality pressure. This study demonstrates that a multivariate predictive index is feasible in rocky reef assessment and illuminates additional investigative work to continue to advance index development.