Eminent domain gives public agencies the right to acquire property from private individuals in exchange for just compensation. Following Kelo vs. New London 2005, the powers of eminent domain were extended, opening the door for potential abuses of power by government officials. In this paper, I empirically investigate the use of eminent domain using newly digitized records from the Tennessee Valley Authority (TVA), which extensively used eminent domain to construct dams in the 1930s. Using a sample of families living in one reservoir location, I find empirical evidence that land tenure, the number of property holdings, and debt levels affected the holdout decision, suggesting that holdouts are individuals with high subjective values. I then develop a screening model with asymmetric information in seller valuations to predict when eminent domain will be used and what settlement values and court awards would be. I test the model's and find that individuals who held out increased their property value by about 5 %, and incentives to holdout existed only for individuals who received initially low settlement proposals.