Heterogeneous expectations and housing market
Date of Issue2017
School of Humanities and Social Sciences
This thesis studies the boom-bust housing price dynamics under heterogeneous expectations. I mainly focus on three key questions: (1) Why some housing prices form an exponential pattern and others follow a linear trend? (2) How to provide an explanation not only for the formation but also for the subsequent collapse of an exponential bubble and a linear trend jointly within the same framework? (3) How do different market conditions affect the trading behavior of market participants in the US housing market? Using Freddie Mac housing price index from the United States, Chapter 1 classifies the dynamics of housing prices on its rising trend into two patterns: an exponential pattern and a linear pattern. An exponential bubble is a housing price bubble that grows exponentially before it bursts. A linear trend is a housing price trend that increases slowly in a linear pattern before it collapses. In Chapter 3, I then develop a three-heterogeneous agent model (HAM) consisting of fundamentalists, chartists and naive agents to show that, without any exogenous shocks, the formations of the two different housing price patterns could have some endogenous origin. To estimate the model, I conduct a nonlinear regression using monthly housing market data in two representative states, Arizona and Minnesota, from Jan 1975 to Jul 2006. The estimated parameters measuring the heterogeneity and switching behavior are statistically significant, confirming the co-existence of the three types of investors and that they are interacting between each other. More importantly, the formation of exponential bubbles and linear trends could be explained by the interactions between heterogeneous agents. Exponential bubbles could be primarily driven by the dominance of the chartists while a linear growth in housing prices could be due to systematically constant proportions of each type of agents. Sensitivity analysis suggests that a housing price movement is more likely to be of an exponential type when we have more frequent transition between trading strategies, stronger belief in the housing trend and smaller gross return on risk-free asset relative to housing. Existing studies only analyze how an exponential bubble and a linear trend are formed or crash in isolation; they rarely look into the formation and the collapse of the two price patterns jointly within the same setup. Moreover, the model in Chapter 3 (BHM-type model) cannot predict the timing of market switches (Bolt et al., 2014). Chapter 4 develops a HAM with an evolutionary selection mechanism based on market-correction-adjusted fitness measure and offers a potential explanation not only for the formation but also for the subsequent collapse of an exponential bubble and a linear trend since the mid of 2006. One important feature of my HAM is that investors not only cluster to the heuristics which enjoyed better performance in the recent past, but also act based on the expectations that the market is more inclined to reverse when the current price deviates further from its fundamentals. Result 1 is that (1) while the formation of an exponential bubble is primarily due to the dominance of chartists, a linear trend can be explained by stable proportions of each type of agents, and (2) the collapse of both an exponential bubble and a linear trend could be driven mainly by the activations of fundamentalists. Result 2 shows that compared to BHM-type (Boswijk et. al., 2007) heterogeneous agent models, our HAM exhibits better forecasting accuracy in terms of predicting the timing of market switches when the dividing point is from Nov 2005 to Feb 2006. The policy implications in Result 3 suggest that a policy maker can stabilize the market through higher interest rate and property tax rate. Chapter 5 proposes a simple heterogeneous agent model with Markov chain regime-dependent beliefs for the U.S. real estate market. The demand for housing is divided in a real component and a speculative component consisting of fundamentalists, chartists and noise traders. Within the Markov-switching framework, the beliefs of chartists and noise traders are modeled to be regime-dependent. I estimate the model using the quarterly housing price data from 1975Q1 to 2015Q11. My Finding 1 is that the model matches well with the boom and bust periods in the U.S. housing market. Finding 2 shows the evidence of time-varying behavioural heterogeneity within-group. To be more specific, chartists form trend following expectation in the boom periods while take contrarian strategy in the bust periods, and noise traders are much more sensitive to external news in the bust state than in the boom state. My finding 3 suggests that the housing price increases (decreases) more drastically when the probability of staying in the boom (bust) state is larger and the bandwagon (contrarian) expectation of chartists is stronger.
DRNTU::Social sciences::Economic theory