Media outlets often bias their news reports. It can be difficult to test the motives behind such biased reporting. We develop models of news provision that detail how the environment and preferences aect demand for, and provision of, news. We examine the implications of these models of media bias, and then test the predictions using weather reports by the New York Times in the late 19th century. This period in history provides a natural experiment for testing bias, as the Times switched from using government weather reports to producing weather reports in-house. When the Times produced its own weather reports it was more accurate at predicting sunny weather and less accurate at predicting rainy weather on days when the local baseball team, the New York Giants, was scheduled to play a home game in Manhattan. The amount of bias is smaller during months that have the least rain, and is higher when the New York Giants are more highly ranked. We find that the empirical results falsify standard neoclassical explanations of media bias. The comparative statics may be better explained by a model of bias with intrinsic preference for information.