Housing Bubbles: Theory, Observations & Synthesis

There are five major questions we need to ask when looking at housing bubbles:

  1. How do housing bubbles get started?
  2. Is it investors or is it normal home buyers who cause a bubble?
  3. How can we tell if a market is overvalued?
  4. How can we tell if the market is about to turn?
  5. What is the probability that prices will fall in the next six months?

The Efficient Market Hypothesis (EMH, Fama 1965, 1970) assumes that current prices fully reflect all public and private information. It contends that market, non-market and inside information is all factored into and that no one has monopolistic access to relevant information. It assumes a perfect market, rational agents, and concludes that prices will then follow a random walk and market excess returns are impossible to achieve consistently. In an efficient market, at any point the actual price of the security will be a good estimate of its intrinsic value. Efficiency, for Fama, means not only that security prices follow a random walk, but also that they wander around their intrinsic or true value. By logical extension, bubbles do not exist if markets are efficient and all agents were rational. The market for housing has some of these attributes.


One of the first critics of the EMH came from Shiller (1981) who pointed to the statistically significant difference of volatility between stock prices (their “assessed” value) and real variables related to their “true” value – e.g. dividends. Several researchers in the late 1990s including Shiller, tried to test to see if housing markets were efficient (Case and Shiller, 1990). The authors conclude that the market for single-family homes did not appear to be efficient since year-over-year home price changes tend to be followed by changes in the same direction in the subsequent year.


Muellbauer and Murphy (1997) argue that some form of credit rationing, and transactions costs all contribute to long run-ups in home prices. Himmelberg et al. (2005) noting the large run-up in home prices in the early 2000s suggested that nonlinearities in the discounting of rents could lead home prices to sharply respond to changes in interest rates in certain markets. Their sample, which extended up to 2004, is the period right before the bust. Glaeser and Gyourko (2006) try to explain the large run up-in house prices using income, amenities and interest rates. In essence, both of these papers argued that fundamental factors could explain the run up in home prices — homebuyers were rational.

However, the 2007/2012 U.S. housing bust, with its large historic price turns in many geographic markets around the nation settled the question for most people. The EMH is not a good descriptor of the U.S. housing market. The 2007/2012 experience showed that homebuyers can display moments of irrational exuberance; housing markets can overheat.

As early as 2003, Shiller suggested there might be a bubble in housing (Shiller 2003). Shiller later (Shiller 2008) argues that failure to value housing based upon its fundamentals combined with future home price optimism lead to feedback-speculative bubbles where home prices are set well above their intrinsic value. Therefore, it is difficult to explain whether house prices in 2006 were too high using a purely rational model.

In subsequent papers, the perfect rationality assumption has been relaxed, allowing the models to shift the focus to how bubbles get started and under

what conditions they would burst. The study of bubbles has expanded to explore the effects of perverse incentives and bounded rationality. Another group of research focuses on market transactions in which one group of agents is irrational.


One substantive idea to come out of the bounded rationality thinking that directly applies to housing is the idea of herding. DeMarzo, Kaniel and Kremer (2008) introduce a relative wealth model, in which an individual agent’s utility depends on not only her absolute wealth but also on her relative wealth (i.e., the agent’s so-called “keeping up with the Joneses” preferences). If that dependency is strong, agents will prefer to participate in bubbles as long as other agents do so, in order to not fall too far behind their peers’ wealth during the bubbles upside.


This herding behavior is often amplified by the popular news media. News stories often focus on assets and liabilities with good past performance. Even if investors might be skeptical that this performance will continue, news stories signal the existence of others who have had great success. Once a market has turned, however, the media is quick to point out the risks (market information is not always information-based). The point of this for someone tracking housing is that swings in home prices can happen very quickly.


A final group of papers comes under the title of rational bubbles. Kermani (2012) proposes a model of booms and bust in housing driven by the interplay between relatively low interest rates and an expansion of credit. Glaeser, Gyourko, and
Saiz (2008) suggested that rational bubbles can exist when the supply of housing is fixed, but not with elastic supply and a finite number of potential homebuyers. Beyond these early works are two dominant strands of thought that are
important to this paper. One relies on heterogeneous beliefs of investors while another relies on heterogeneous time horizons of investors. The first strand of rational bubbles based on investor behavior posits that investors could have
access to all the information that other investors have, but each investor is differentiated because they have different prior belief distributions possibly due to psychological biases. In such a case, investors with non-common priors can agree to disagree even after they share all their information. An attractive feature of heterogeneous-belief bubbles is that they predict that bubbles are accompanied by large trading volume and high price volatility, as for example in the model by Scheinkman and Xiong (2003). DeFusco, Nathanson, and Zwick (2017) also
point to the role of speculators as the cause of asset bubbles. Investors expect prices to increase after past increases. These authors, however, argue that investors can have homogenous beliefs based upon common information yet have heterogeneous horizons regarding how long they will hold a property. Investors with shorter time horizons tend to pull out of the housing cycle before owners who plan to continuously occupy a purchased property. So for DeFusco et al. (2017), it is neither ordinary homebuyers who irrationally bid up home prices using a home as investment good, nor investors who differ in their beliefs. The run up in prices is due to speculative investors with short time horizons who do not expect to live in the home. They find that a sharp rise in non-occupant purchases explains much of the variation in purchase volume across and within CBSAs between 2000 and 2005.


This focus on investors who rationally enter and exit a market, however, ignores the ordinary homebuyer. For the ordinary homeowner, even though housing functions as a source of shelter it also functions as an investment. If there were no investment role to housing, bubbles would be less likely. This is because there would be less incentive to buy more house than one would need. Real home prices would rise roughly with the cost of materials and/or the ability of a geographic area to attract in-migration, or faster than other areas mainly due to it having a better job market or a better climate. Housing, however, does have an investment role. The ability to borrow money, to leverage one’s down payment, has made it a powerful asset to make, or lose, money for its owners. These buyers are both the people who intend to live in the home, as well as those who are speculative buyers. These ordinary home buyers, this paper argues, can be irrational.


The asset pricing literature long ago showed how difficult it is to confirm the presence of a bubble (e. g., Flood and Hodrick, 1990). More recently, however, Phillips, Shi and Yu (2015) show that housing bubbles can be tested for and time stamped using a right-tailed augmented Dicky Fuller test. Their approach is one that measures if homebuyers are paying considerably more for a house over time than suggested by its intrinsic value (rent) or its equilibrium value (income) — stationarity has been violated. We adopt this theoretical view, but use a more heuristic model. The ability to determine if a market is overvalued depends upon accepting that buyers can display periods of irrational exuberance and that one has the ability to establish this at any point in time by looking at the cost of owning versus the cost renting close substitutes. Rent in a sense measures fundamental or intrinsic value. In our framework, a housing market can overheat when potential homebuyers (or homeowners), as a collective group, have unreasonably high expectations about future capital gains, and they pay significantly more to own a house than to rent a similar property. So, for this paper, the cost of owning not being close to the cost of renting is a telltale sign of a bubble, ex ante. Given this, how can the practitioner determine overvaluation?



2.1 How can we measure if a market is overvalued?


According to Minsky (1982) and Kindleberger (2000) there are three kinds of bubbles: ones that go up and then crash suddenly and more or less totally, ones that go up and then go back down again without a hard crash, and ones that go up and reach a peak and then go into a period of financial distress and then later crash. The vast majority are the latter. Oil followed the first pattern in 2008, peaking at $147/barrel in June 2008, then plunging to the low 30’s by November 2008 before recovering. Real estate in the early 1990’s followed the second pattern, going back down the way it went up. This was because residential homeowners have an emotional investment in their houses and refuse to believe or accept the falling prices after the peak. So they do not sell, and volumes collapse rather than prices.


Moreover, an overvalued housing market could unwind without popping. It could do this by incomes simply inflating over a long period of time (essentially, incomes, population or rents would rise faster than prices). It is also interesting to think that an overvalued housing market like those that we currently witness in Canada, New Zealand and Australia could remain in disequilibrium indefinitely, with the cost of owning much higher than the cost of renting, as long as there is trade exchange imbalance and strong capital inflows from an external source (China). Houses are places where people park money. These markets would remain overvalued over the long run. And it is true that an overvalued housing market could quickly crash in the first two ways Kindleberger points out above and we might not detect its turning.


However, it is equally likely that a boom/bust evolves the third way Kindleberger suggested (in three stages) — it goes up and reaches a peak, goes into a period of financial distress and then later crashes). In this evolution, the build-up to the peak could last a long time. If this is/were the case (a building up to the peak – the run-up phase) we practitioners should at least be aware that a market has become overvalued and that there is a probability that it will turn.


2.1.a Vector error correction models with Pt > Pt* (VECMs)


In 2007Q2, the median home price for the nation fell for the first time in recently recorded history. It then fell for the next 19 consecutive quarters for a cumulative decline of 23 percent. How did we miss it? Actually, there were debates and several research papers written during the buildup which contained evidence suggesting the magnitude of the potential price declines using a two-stage fitted home price index approach. As early as 2003, Shiller suggested there might be a bubble in housing. Gallin (2004, 2006) using a user cost and rent approach argued that housing was overvalued. Gao, Lin and Na (2009) using a two stage vector error correction model (VECM) also suggested that many markets were dramatically overvalued. The two stage VECM models rely on forecasting an equilibrium home price Pt* in the first stage and comparing it to actual prices Pt in stage 2. If Pt > Pt* then a market is overvalued. The problem is that knowing that a market is overvalued is a necessary condition, but not a sufficient condition to say that a strong downturn is going to happen next. The reason is that an overvalued market could unwind in several ways.


2.1.b Trading volume by investors with short time horizon is high (STIs)


A second approach to determine if a market is overvalued relies on investors with short time horizons. In these models, some buyers plan to sell after one year, while others plan to hold for many years. DeFusco et al. (2017) specify a term structure for extrapolation. They model a housing market populated by investors with extrapolative heterogeneous horizons. They believe that extrapolation declines with the forecast horizon so that short-run expectations display more sensitivity than long-run expectations. Importantly, they find that the marginal effect of short-term potential buyers on the price and the volume is quantitatively large and find that a sharp rise in the non-occupant purchases explains much of the variation in volume across CBSAs and across time between 2000 and 2005.


For these authors, homebuyers have extrapolative expectations, not irrational expectations – they expect prices to go up after past increases, but no faster than past increases. As speculators they do not look at intrinsic value (e.g., rent). So it is neither homebuyers who intend to live in the property nor property investors planning to rent the property who cause a market to become mispriced, it is purchases by speculative buyers with short horizons. A telltale sign of a bubble is thus when there is a rising share of speculative buyers. For these authors, volume peaks well before prices and their ability to measure the rise in volume is an indicator that the market is overheating.