Forecasting Real Housing Price Returns of the United States using Machine Learning: The Role of Climate Risks

Climate change, a pressing global challenge, has wide-ranging implications for various aspects of our lives, including housing prices. This paper delves into the intricate relationship between climate change and real housing price returns in the United States, leveraging a comprehensive dataset and advanced machine learning technique - the step-wise boosting method. This sophisticated ensemble learning technique, known for its iterative refinement process that emphasizes correcting errors made by previous models, significantly enhances our analysis. By strategically focusing on data points that previous iterations have misclassified and minimizing the exponential loss function, step-wise boosting allows for a nuanced understanding of how climate variables affect housing prices. Our findings suggest that climate change variables can influence real housing price returns, particularly in the short term, but the relationship is complex and varies by region. The adaptive learning capability of step-wise boosting, which meticulously adjusts the weights of incorrectly classified instances to improve accuracy and learning efficiency, has been crucial in uncovering these insights. This methodological approach not only underscores the importance of employing advanced predictive models in analyzing the effects of climate change on urban development but also highlights the potential for informed decision-making, sustainable urban planning, and climate risk mitigation.