While Fang and Peress (2009) discover the media coverage premium in the U.S. market, we find that this anomaly also exists in the Chinese stock market. We further classify the media articles into positive, neutral, and negative sentiment to test the media sentiment anomaly in the cross-sectional stock returns. We find that firms without positive news or negative news have significant positive excepted returns. Portfolios constructed by longing firms without negative news and shorting firms with high negative news could obtain the highest abnormal return after considering well-known risk factors. Moreover, we test the different combinations of portfolio formation and holding period. We find that the excess return exists for more than three months holding periods with different portfolio formation periods by news sentiment. We conclude that news sentiment anomalies are not caused by the short-term momentum effect and are robust in the Chinese stock market.