We study the long-term impact of limited attention on belief formation. To this end, we propose a rank-dependent model of inattentive learning, building on principles from the behavioral science literature. We provide explicit formulas for the distorted limiting beliefs and the asymptotic variance that characterizes the speed of learning. In our model, agents who pay excessive attention to extreme observations end up learning a distorted version of the true data-generating process which puts too much weight on the tails of the distribution. We show how the implications of our model can be linked to empirically documented phenomena such as skewness-seeking behavior and the coexistence of over- and underreaction to incoming information depending on the context. Limited attention can thus have long-lasting impacts on the beliefs agents form.

I document a new empirical regularity in initial public offerings (IPOs): trading volume in the 600 business days following an IPO follows a U-shaped pattern. While the initial surge in turnover is well-established, I also show that trading activity subsequently rises again, with average volume increasing by approximately 22\% per annum relative to the post-listing trough. This phenomenon has not previously been reported. The increase is robust to aggregate market turnover trends and to firm size at issuance. I further demonstrate that the U-shape is linked to lockup agreements, yet not driven by insider sales or free float.