Standard neoclassical models in finance assume that individuals form expectations and make decisions using all available information. While these theories dictate that new information is instantaneously incorporated into asset prices, our minds and cognitive resources are finite and we allocate our attention selectively. Moreover, the amount of information relevant to the valuation of an asset is far from trivial in the current information society. This has important implications for asset pricing because attention is a prerequisite for distilling and processing information into prices. The underlying motivation of the empirical essays in this dissertation is to better understand the implications of boundedly rational market participants and to examine the role of limited attention. Using innovative datasets to measure attention, it provides new insights into how limited attention affects the expectation formation and behavior of financial agents, such as investors and analysts, and how this ultimately feeds into asset price dynamics. In this way, this dissertation contributes to a further development of the finance discipline from its neoclassical foundations towards a more realistic approach that integrates behavioral phenomena. In the first chapter we empirically test the rational inattention model for exchange rates. Rational inattention theory provides a framework of how cognitively limited agents might simplify and summarize available information. The framework is particularly suitable to study the relationship between exchange rates and fundamentals, because exchange rates are largely determined by expectations of market participants. Our results provide strong evidence in favor of the rational inattention theory of exchange rates.