The present study considers factors that motivate users of social networks to publish different types of privacy-related information to friends or even the public. In contrast to prior research, we do not limit our research scope to an individual's decision-making process (i.e., the formation of behavioral intentions) but also include actual behavior as observed among a group of real Facebook users. Our objective is to test to what extent existing theory is not only capable of explaining self-disclosure decisions but also to predict subsequent behavior. We test our model using a combination of structural equation modeling and logistic regression with questionnaire data and data collected from the Facebook platform. Our results indicate that the way self-disclosure was operationalized in prior research shows low predictive power, especially when compared to predictions based on simple questions regarding an individual's sensitivity to the disclosure of personal information.