Arrested Development: Theory and Evidence of Supply‐Side Speculation in the Housing Market

• Date: 01 December 2018
• Author: CHARLES G. NATHANSON,ERIC ZWICK
• DOI: http://doi.org/10.1111/jofi.12719
• Published date: 01 December 2018

THE JOURNAL OF FINANCE •VOL. LXXIII, NO. 6 •DECEMBER 2018

Arrested Development: Theory and Evidence of

Supply-Side Speculation in the Housing Market

CHARLES G. NATHANSON and ERIC ZWICK∗

ABSTRACT

This paper studies the role of disagreement in amplifying housing cycles. Speculation

is easier in the land market than in the housing market due to frictions that make

renting less efficient than owner-occupancy. As a result, undeveloped land facilitates

construction and intensifies the speculation that causes booms and busts in house

prices. This observation challenges the standard intuition that in cities where con-

struction is easier, house price booms are smaller.It can also explain why the largest

house price booms in the United States between 2000 and 2006 occurred in areas

with elastic housing supply.

ASSET PRICES GO THROUGH PERIODS of sustained price increases, followed by

busts. To explain these episodes, economists have developed theories based on

disagreement, speculation, and strategic trading. This literature focuses on the

behavior of asset prices in stock markets. But do these ideas explain housing

markets as well? Like any other financial asset, housing is a traded, durable

claim on uncertain dividends. Moreover, an enduring feature of housing mar-

kets is booms and busts in prices that coincide with widespread disagreement

about fundamentals (Shiller (2005)), and there is a long history of investors

using real estate to speculate about the economy (Kindleberger (1978), Glaeser

(2013)).

In this paper, we incorporate disagreement into a housing model to exam-

ine whether the finding that disagreement raises stock prices generalizes to

∗Charles G. Nathanson is at Northwestern University and Eric Zwick is at the University of

Chicago and NBER. Wethank John Campbell, Edward Glaeser, David Laibson, and Andrei Shleifer

for outstanding advice and Bruno Biais (the Editor), Tom Davidoff, Morris Davis, Robin Green-

wood, Sam Hanson, Mitchell Hurwitz, Chris Mayer, Andrew Paciorek, Alp Simsek, Jeremy Stein,

Amir Sufi, Adi Sunderam, Stijn Van Nieuwerburgh, Paul Willen, and two anonymous referees for

helpful comments. We also thank Harry Lourimore, Joe Restrepo, Hubble Smith, Jon Wardlaw,

Anna Wharton, and CoStar employees for enlightening conversations and data. Prab Upadrashta

provided excellent research assistance. Nathanson thanks the following organizations (none of

which reviewed this paper) for financial support: the NSF Graduate Research Fellowship Pro-

gram, the Becker Friedman Institute at the University of Chicago, the Guthrie Center for Real

Estate Research, and the Bradley Foundation, a group that supports limited government, a topic

tangentially related to this paper.Zwick thanks the University of Chicago Booth School of Business,

the Neubauer Family Foundation, and the Harvard Business School Doctoral Office for financial

support. We have read the Journal of Finance’s disclosure policy and have no further conflicts of

interest to disclose.

DOI: 10.1111/jofi.12719

2587

2588 The Journal of Finance R

the housing market.1Housing differs from the typical asset studied in finance

in two fundamental ways. First, the underlying real assets in many financial

models are in fixed supply.2Incontrast, elastic supply is central to housing mar-

kets, as firms respond to high prices with new construction (Gyourko (2009)).

Second, while the typical financial asset pays cash dividends, the dividend paid

by housing is the flow utility enjoyed by end users. This flow utility has differ-

ent values for different people and is not perfectly transferable, leading many

people to prefer owning to renting (Henderson and Ioannides (1983)).

We study a two-period model of a housing market with two classes of agents.

Potential residents receive heterogeneous utility from consuming housing that

accrues only when they own their houses. Developers supply housing in a com-

petitive market, buying land at market prices, and converting it into housing

for a constant resource cost. As in Saiz (2010), the amount of developable land

is fixed due to geographic and regulatory constraints. At t=0, there is an

initial number of potential residents N0,andatt=1, the number of potential

residents grows to N1=eμxN0, where xis a positive shock. The larger is N1,the

greater are the price of land and housing at t=1. At t=0, all agents observe

the shock x, but they “agree to disagree” about μ. In this context, we define a

“house price boom” as the reaction of the house price at t=0totheshockx.

Our results characterize how the size of the house price boom varies with

N0.WhenN0is very small, there is no house price boom because the price

of land remains equal to zero. When N0is very large, initial housing demand

can be so strong that, at t=0, all space is held by homeowners. In this case,

the house price at t=0 reflects both heterogeneous expectations about the

t=1 price and homeowners’ flow utility for housing. The latter mitigates the

impact of the former. Finally, when N0is intermediate, some land at t=0is

left unoccupied by homeowners and is held by developers. These supply-side

speculators hold land only when they have optimistic beliefs about μ, and it is

these optimistic beliefs that drive prices. Consequently,prices at t=0 are most

sensitive to disagreement and optimism for cities with an intermediate level of

development. For the same reason, the house price at t=0 is most sensitive to

x, that is, booms are larger, in these cities.

Stated in terms of available land in the city, the house price boom is largest

for intermediate values of initial land supply. This nonmonotonicity between

the house price boom and supply contrasts with prior work on disagreement—

which does not consider the unique aspects of housing—and prior work on

1Beginning with Miller (1977), a large literature uses models of disagreement to explain asset

pricing patterns in the stock market. Hong and Stein (2007) survey this literature, which includes

Harrison and Kreps (1978), Morris (1996), Diether, Malloy, and Scherbina (2002), Scheinkman

and Xiong (2003), Hong, Scheinkman, and Xiong (2006), and Simsek (2013). Other papers apply

speculative finance models to housing. See, for example, Piazzesi and Schneider (2009), Favara

and Song (2014), Giglio, Maggiori, and Stroebel (2014), and Burnside, Eichenbaum, and Rebelo

(2015). Unlike those papers, our work focuses on housing supply.

2Although firms can (and do) issue new equity in response to high demand for stock and high

prices (Baker and Wurgler (2002)), many asset pricing models assume that the underlying real

assets of a firm are fixed. See, for instance, Lucas (1978) or Scheinkman and Xiong (2003).

Arrested Development 2589

housing cycles—which does not consider disagreement. Taken separately, each

approach predicts a monotonically declining relationship between a house price

boom and initial land supply (Hong, Scheinkman, and Xiong (2006), Glaeser,

Gyourko, and Saiz (2008), Paciorek (2013)).3By joining these approaches, we

provide a new insight about housing cycles that neither approach offers alone.

We demonstrate the robustness of this result in several extensions that relax

the model’s assumptions in different ways. First, we consider the case in which

developers can issue equity and investors can short-sell that equity.Second, we

consider an extension in which landlords can speculate in the housing market

and rent housing to pessimistic residents. In a final extension, we generalize the

model to the case in which supply elasticity declines continuously with the level

of initial demand. We also formally show how disagreement reduces welfare

(in the sense of Brunnermeier, Simsek, and Xiong (2014)) by reallocating space

from high-flow-utility pessimists to low-flow-utility optimists and developers.

The model’s core insight helps explain the variation in 2000 to 2006 house

price booms across U.S. cities. As shown by Davidoff (2013), a static supply-

demand framework with a common national demand shock cannot account

for cities in the “sand states” of Arizona, California, Florida, and Nevada that

experienced strong price and quantity growth.4One possibility is that the de-

mand shocks in these cities were especially large due to local credit conditions,

differences in productivity and amenities, or heightened speculative activity

by homebuyers (Barlevy and Fisher (2011), Davidoff (2013), Gao, Sockin, and

Xiong (2016)). Our model offers an alternative explanation: house prices were

more sensitive to a demand shock in these cities because the cities were at an

intermediate level of development, and market participants disagreed about

future prices.

We offer several pieces of empirical evidence in support of this explanation.

Land price growth from 2000 to 2006 was high and closely correlated with house

price growth across cities, whereas construction cost growth was not. Matching

this fact distinguishes our model from Gao, Sockin, and Xiong (2015), who offer

a theory of nonmonotonic house price booms that is agnostic on the differential

roles of land prices and construction costs. We also document land market

speculation from U.S. public home builders. These firms amassed land far in

excess of their immediate construction needs during the boom, while investors

short-sold home builder stocks more than nearly every other industry. The

model predicts these outcomes for developers in intermediate cities only in the

presence of disagreement.

Although we do not systematically examine land constraints in sand state

cities, the case of Las Vegas offers a stark illustration of our model. Las Vegas

is surrounded by land owned by the federal government, and Congress passed

3Simsek (2013) predicts that the introduction of new nonredundant financial assets increases

the “speculative variance” of portfolio values, but this result does not apply to our findings because

the price of undeveloped land is perfectly correlated with the price of housing in our equilibrium.

4Understanding the factors driving these outlier cities is crucial for evaluating research in-

strumenting for price growth with supply elasticity. Davidoff (2016) discusses problems with this

instrument.