A Data-Centric Approach to Post-Pandemic Housing Market Analysis

The housing market in California has gone through some unexpected fluctuations during the COVID-19 pandemic. The housing inventory reached at all-time lows and prices reached highs not seen since before the economic crisis of mid2000s. This paper explores these trends using Central California as a case study. Home valuation on home sales since 2018 in Fresno County of Central California are modeled with geographically weighted regression, geographically and temporally weighted regression, as well as artificial neural networks. Moreover, autoregressive integrated moving average models are utilized to explore the trends of price and days on the market in Fresno County. Finally, statewide COVID-19 data and Blueprint for a Safer Economy are used to predict changes in year over year housing prices across various California counties. The geographically weighted regression model outperformed the other two valuation models, both of which incorporate the date a home was sold as a predictor. The time-series analysis on Fresno County trends demonstrates unpredictability during the pandemic. The neural network underperformed in predicting sales price in counties, indicating that different policy and outbreak status of certain counties may not have an effect on the increasing home prices in California.