According to the privacy calculus, both privacy concerns and expected gratifications explain self-disclosure online. So far, however, most findings were based on self-reports, and little is known about whether the privacy calculus can be used to explain observations of actual behavior. Likewise, we still know little as to whether the privacy calculus is influenced by the design of online websites, including for example popularity cues such as like and dislike buttons. To answer these questions, we ran a preregistered one-week field experiment. Participants were randomly distributed to three different websites, on which they discussed a current political topic. The websites featured either (a) like buttons, (b) like and dislike button, or (c) no like/dislike buttons, and were otherwise identical. The final sample consisted of 590 participants. Although the originally preregistered model was rejected, the results showed that a considerable share of actual self-disclosure could be explained by privacy concerns, gratifications, privacy deliberation, trust, and self-efficacy. The impact of the popularity cues on self-disclosure and the privacy calculus was negligible.