EXPLORING ELEMENTS DRIVING THE BARGAINING POWER OF LANDED RESIDENTIAL PROPERTY IN SINGAPORE

Abstract

This thesis paper explores the drivers behind bargaining power in landed residential property in Singapore. Three machine learning models namely Random Forest, XGBoost and CatBoost are applied in comparison to traditional Ordinary Least Squares (OLS) Hedonic regression. Results found that bargaining power cannot be quantified when it is defined as a ratio of transacted price to listed price – ??1. This could be because negotiation between two parties is a complex process with many factors coming into play. Therefore, it is not possible to capture nor predict the idiosyncrasies and complex nature of human behaviour in a quantitative way as it may vary on a case by case basis. When bargaining power is defined as being relative to other transacted prices – ??2, it is found that the CatBoost model performed the best with a relatively high ??2 of 89.4% , identifying tenure and purchaser address indicator as key contributing drivers. Secondary findings include the superior performance of machine learning methods as compared to traditional OLS hedonic regression. Therefore, machine learning methods can be considered as an alternative or used to complement findings from traditional hedonic regression.