This dissertation contains three chapters related to housing markets. First two chapters study the vertical variation in housing prices, and the third chapter models buy-sellers. The first focus is the vertical dimension of housing markets. Location is the most important factor in determining housing prices. However, in multi-unit residential markets, two housing units in one building share the same location but may still have different transaction prices. The first chapter is to explore why. To identify the explanatory factors, the author collects over 200,000 transaction records from 2012-2016 in Beijing and provides rich and valid empirical evidence. Also, this chapter proposes a least nugget effects estimator (LNE) as an alternative to the conventional fixed effects estimator (FE). LNE is essentially a pairwise differences estimator (PD) whose comparison with FE in linear fixed effects models is studied in the second chapter. The second focus of this dissertation is a type of households who are both home buyer and home seller at the same time (called "buy-sellers"). The traditional framework would model such households as two independent identities. However, this dichotomic tradition ignores the essential feature of the buy-sellers, that is, the interdependence between buying and selling decisions. As a buyer, a new purchase depends on a successful sale due to budget and policy restrictions; as a seller, a sale may rely on the success in bidding to avoid renting for a living. This interdependence results in a connected bidding network. One’s success or failure in bidding or selling will influence others’. Chapter three, for the first time, studies this type of households and proposes an agent-based model (ABM) with buy-sellers. Then, the author uses this ABM to analyze the effects of housing purchase restrictions and home brokerage on market outcomes.