This thesis is comprised of three essays. The first two are about how firms make investment decisions and their consequences on macroeconomics. The last one is on the investors behavioral biases in the Non-Fungible Token market.The first essay is co-authored with Zebang Xu. We argue that accounting for capital heterogeneity is important for understanding the sources and costs of misallocation. We prove that, conditional on the same observables, further disaggregation of capital types will always lead to a higher measured cost of misallocation. Quantitatively, accounting for capital heterogeneity increases measured costs of misallocation by 7 p.p. (19%) in the U.S. and 6 p.p. (24%) in India. Across countries, structures are consistently more misallocated than equipment. We then estimate a dynamic model to disentangle sources of misallocation that can explain the additional measured misallocation and the efficiency differences between equipment and structures. Results indicate that adjustment costs and imperfect information cannot fully explain the additional misallocation or why structures are more misallocated. Heterogeneous financial constraints and tax policies may contribute to the higher misallocation of structures, while heterogeneous technology and measurement errors play only modest roles. The second essay explores how belief shocks affect firms’ investment decisions through different channels of their expectations. The literature has well documented the causal effect of belief shocks on firms’ behavior, but it is lessclear through which expectations these belief shocks exert their influence. I develop a stylized firm investment model to demonstrate the mechanism by which a belief shock affects investment through various types of expectations. Using constructed belief shocks and the Duke CFO Survey, I document the heterogeneous effects of belief shocks on different expectations. By linking firm expectations with fundamental datasets, I also measure the relative importance of these expectations in explaining firms’ investments and efficiency. Finally, I construct an overarching expectation index that captures the main content of firms’ different expectations using Principal Component Analysis. Estimating the causal effect of expectations on firms’ investment can yield significantly different results when using the index compared to a single expectation. The third essay, co-authored with Lin William Cong and Xiangchen Liu, studies anchoring effects and loss aversion in the Non-Fungible Token market using comprehensive user-auction data from CryptoPunks. Using the last period’s transaction price as the reference point, we separately measure anchoring effects and loss aversion for sellers and bidders. Our findings reveal that both sellers and bidders exhibit loss aversion, though the magnitude of loss aversion is significantly smaller for bidders. The analysis also confirms the presence of anchoring effects in both groups, with bidders displaying weaker effects than sellers, possibly due to signaling dynamics. Furthermore, we show that negotiations between sellers and bidders result in an intermediate level of loss aversion, falling between their individual measured levels. We also test whether experience influences loss aversion within both groups. For sellers, greater experience mitigates loss aversion, whereas for bidders, the results are mixed, with some evidence suggesting that experience may even increase their loss aversion in the NFT market.