Description:

Speaker: Kate Hauck

Title: Dynamic Learning over Beliefs about Farming in the American West

Abstract:
Many settings involve dynamic learning and its consequences, including farmers learning about fertilizer in developing countries, students learning about the returns to college majors, and firms learning about the demand for their products. To reflect this dynamic learning, we extend a popular dynamic optimization model to allow agents to change their information sets over time and to update their beliefs. In particular, we allow for dynamic optimization and unobserved differences between agent types, and we allow these types to have biased Bayesian beliefs. We apply this model to the setting of individual land acquisition in the late nineteenth century, when the U.S. government granted over 11% of the land area in the country to homesteaders. Individuals who homesteaded were required to farm the land. Thus, these individuals learned about the value of farming. They used this updated belief about the value of farming to decide whether to (i) acquire the title to the farm by farming for five years, (ii) abandon the farm, or (iii) sell the farm. We digitize and match novel plot-level histories of land acquisition, land resale, and agricultural production to estimate the farmers' learning process. We use our model to construct counterfactuals that show the impact of individuals' beliefs on their decisions. Specifically, 62% of the farmers who abandoned their farms within the first five years would not have done so if they had begun with the beliefs they would have held after ten years of farming. In a second counterfactual, we demonstrate that if the U.S. government had not offered homesteading, only 31% of the homesteaders would have purchased the land instead and 69\% would have opted not to farm. These results indicate that without the Homestead Act, the speed of western expansion would have been reduced by 38%.