We use a novel database to study the performance of real assets created at different points of an investment cycle. Our database contains information on the construction year of virtually all hotels in the U.S., and performance information for the majority of them from 2000 to 2009. This allows us to identify the impact of hotel investment booms on the performance of hotels built during those booms. We find that, controlling for local (county-level) construction cycles, aggregate (U.S. level) booms do not have a long-term effect. However, hotels that were built during local hotel investment booms underperform their peers over several decades. We examine possible explanations for this long-lasting underperformance. We find evidence consistent with the presence of herding. That is, potential entrants in a local market regard the construction of a hotel as good news about that market and are more likely to enter themselves, but are later disappointed.