Anchoring and Asymmetric Information in the Real Estate Market: A Machine Learning Approach
Journal of Risk and Financial Management, 14(9), 423. https://doi.org/10.3390/jrfm14090423
The University of Auckland Business School Research Paper Series
Posted: 30 Nov 2021
Date Written: 2021
Abstract
Conventional wisdom suggests that non-local buyers usually pay a premium for home purchases. While the standard contract theory predicts that non-local buyers may pay such a price premium because of the higher cost of gathering information, behavioral economists argue that the premium is due to buyer anchoring biases in relation to the information. Both theories support such a price premium proposition, but the empirical evidence is mixed. In this study, we revisit this conundrum and put forward a critical test of these two alternative hypotheses using a large-scale housing transaction dataset from Hong Kong. A novel machine-learning algorithm with the latest technique in natural language processing where applicable to multi-languages is developed for identifying non-local Mainland Chinese buyers and sellers. Using the repeat-sales method that avoids omitted variable biases, non-local buyers (sellers) are found to buy (sell) at a higher (lower) price than their local counterparts. Taking advantage of a policy change in transaction tax specific to non-local buyers as a quasi-experiment and utilizing the local buyers as counterfactuals, we found that the non-local price premium switches to a discount after the policy intervention. The result implies that the hypothesis of anchoring biases is dominant. Full paper available at https://doi.org/10.3390/jrfm14090423
Keywords: unsupervised machine learning, natural language process, non-local buyers, anchoring biases, information asymmetry, repeat-sales estimates
Suggested Citation: Suggested Citation