The Impact of Media Type on Belief Variance: Evidence from a Panel Study

This paper examines how traditional and nontraditional media influence people's beliefs about future states, focusing on the variance of these beliefs. Nontraditional media, such as Internet search engines, allow individuals to select information according to their interest more easily than traditional media, such as television and newspapers, which provide relatively homogeneous information. Consequently, the media type can affect not only the mean but also the variance of beliefs, resulting in greater variance among nontraditional media users compared to traditional media users. We utilize a unique panel dataset that asks respondents about their main sources of COVID-19-related information and their predictions about the end of the pandemic. Since the prediction variable in our data is categorical, we apply an interval censored fixed effects regression model. The estimation results show that the variance of the predictions is significantly smaller for traditional media users than for nontraditional media users. At the population average, the standard deviation is almost one month larger for nontraditional media users, leading to a prediction interval that is approximately 20% wider. Further analysis suggests that information sources predominantly influence infection prevention behaviors through their impact on subjective beliefs about the pandemic's end rather than perceptions of the disease risk.