The present value model states that the price of an asset is equal to the properly discounted future cash flows generated by this piece of asset. This view needs to be modified in real estate markets. Agents in a real estate market can be broadly divided into those who buy to use and those who buy to lend. If rents are too high, occupants may consider buying instead of renting, and vise versa. This implies that there are dynamic interactions between the rental and the price of a real estate. In particular, if there is a speculative bubble in the price, that bubble will also be transmitted into the rental. Our empirical analyses using data from city office show that our dynamic structural simultaneous equation model improves significantly over the single equation present value model. The result is further improved by allowing model parameters to shift value according to a latent variable which follows a first-order Markov chain.