Using a worldwide sample of 3250 global equity funds, we provide out-of-sample evidence of active share as a strong return predictor. However, a global fund’s within-region active share predicts superior performance in Europe and Asia-Pacific, but not in the United States. We reconcile this difference by showing that highly active global managers (whether based in the U.S. or elsewhere) have outperformed both in U.S. and international markets primarily when they are also betting on equity anomalies. The weak return predictability of active share alone in the U.S. stems from domestic anomalies and is not generalizable to global markets.

We introduce a new holdings-based procedure to identify whether a mutual fund has a benchmark discrepancy, which we define as a benchmark other than the prospectus benchmark best matching a fund’s investment strategy. We find that funds with a benchmark discrepancy tend to be riskier than their prospectus benchmarks indicate. As a result, the funds on average outperform their prospectus benchmarks, before further risk adjustments, despite underperforming the benchmarks that best match their portfolios.

We test the biasedness of unsolicited ratings relative to solicited ratings using the ex post firm performance measured by the long-run stock performance of firms following rating announcements and changes. We find that the announcements of new unsolicited ratings are followed by negative long-run stock performance, while those of new solicited ratings are followed by insignificant long-run stock performance. These results are inconsistent with the conservatism hypothesis that suggests that unsolicited ratings are downward biased. We further demonstrate that firms with solicited upgraded (downgraded) ratings experience subsequent positive (negative) abnormal stock performance, while those with unsolicited rating changes have zero abnormal stock performance. The differential stock performance following rating changes between solicited and unsolicited ratings reflect the differential information carried by each type of rating rather than the biasedness in ratings. Specifically, while solicited ratings are based on both public and private information, unsolicited ratings are mainly based on public information. Overall, we find no evidence for a downward bias in unsolicited ratings.