The Political Economy of Housing Supply:
Homeowners, Workers, and Voters
University of Wisconsin–Madison
London School of Economics
Discussion Paper No. TE/2007/512
The Suntory-Toyota International Centers for
Economics and Related Disciplines
London School of Economics and Political Science
London WC2A 2AE
Address for correspondence: François Ortalo-Magné, School of Business, University of Wis-
consin –Madison, 975 University Ave, Madison, WI 53706, USA. email@example.com.
We are grateful to Morris Davis, Dwight Ja¤ee, Erzo G.J. Luttmer, Stephen Malpezzi, Monika Piazzesi,
John Quigley, Silvana Tenreyro, Branko Urosevic, seminar participants at the Bureau of Labor Statistics,
New York University, North Carolina State University, Université de Toulouse, University College London,
University of Birmingham, University of California at Berkeley, University of Manchester, University of
Pennsylvania, University of Southern California, University of Wisconsin–Madison, Warwick University, and
conference participants at the 2007 Winter meetings of the Econometric Society, the 2007 Annual meeting of
the AREUEA, the 2006 NBER Summer Institute, and the 2006 North-American RSAI conference for useful
comments and suggestions.
Equilibrium of the housing market depends on a complex set of interactions between: (1)
individual location decisions; (2) individual housing investment; (3) collective decisions on
urban growth. We embed these three elements in a model of a dynamic economy with two
sources of friction: ill-de…ned property rights on future land development and uninsurable
shocks a¤ecting labor productivity. We characterize the feedback between the households’
desire to invest in housing as a hedge against the risk of rent ‡uctuations and their support
for supply restrictions once they own housing. The model generates an ine¢ ciently low
supply of housing in equilibrium. The model also rationalizes the persistence of housing
undersupply: the more restricted the initial housing supply, the smaller the city size selected
by the voting process. We use the model to study the e¤ects of a number of policies and
Keywords: Housing Supply, Housing Demand, Regulatory Policies, Political Economy.
JEL Nos.: R31, R21, R38, D72.
c The authors. All rights reserved. Short sections of text, not to exceed two paragraphs,
may be quoted without explicit permission, provided that full credit, including c notice, is
given to the source.
An increasing body of evidence points to the importance of supply restrictions in under-
standing housing price dynamics. Glaeser, Gyourko and Saks (2005a) report that changes
in regulatory regimes explain the scarcity of land for housing development in what are today
the most expensive U.S. housing markets. They point to di¤erences in man-made scarcity
as a determinant factor for the explosion of the dispersion in housing prices across U.S.
housing markets since the mid-seventies. Quigley and Raphael (2005) also blame housing
regulation for the recent housing price boom in California.1 Green, Malpezzi and Mayo
(2005) …nd that housing supply regulations are the key driver of di¤erences in housing sup-
ply elasticities across U.S. metropolitan areas. In the United Kingdom, Barker (2003, 2005)
identi…es the regulatory constraints on the release of land for housing development as the
primary reason behind the unresponsiveness of housing supply to price increases.
What are the determinants of housing supply regulations? Restrictive supply regulations
cannot survive without political support. To understand the political economy of housing
supply, we need to understand who participates in the decision process, the stakes of the
participants, and the mechanism whereby participants’ preferences translate into policies.
We then need, at a minimum, a location choice model to determine who lives in a particular
area and a housing investment model to predict what real estate assets the residents own.
We also need a collective choice model to map the identities and preferences of local residents
into political decisions over urban growth.
Our goal in this paper is to provide a …rst step towards a theory that encompasses
these three elements: housing consumption, housing investment and collective choice over
housing supply regulation. Each of these three elements is quite complex and linked to the
others through multiple channels. We do not try to capture in one model the richness of
the institutions that regulate these three phenomena. Instead, we o¤er a parsimonious and
tractable framework to gain some insight into the basic issues and to link areas of research
that have traditionally been separate.
In this spirit, we assume only two deviations from complete markets. The …rst is a
key feature of housing markets: building permits are needed for new construction, and
they are issued by the local government. Hence, while property rights on existing buildings
are relatively well-de…ned, property rights on future construction are blurred. The second
1 For further evidence on the critical impact of housing regulations on housing supply, see Ozanne and
Thibodeau (1983), Rose (1989), Malpezzi, Chun and Green (1998), Mayer and Somerville (2000), Glaeser
and Gyourko (2003), Glaeser and Ward (2006), Glaeser, Gyourko and Saks (2005c).
deviation is a staple assumption in macroeconomics: Households cannot insure against
future labor income shocks. As we shall see, the combination of these two imperfections
is su¢ cient in equilibrium to generate an undersupply of housing and cause persistence of
In the baseline model, we consider a country with one city and a vast countryside.2
A continuum of agents live for two periods. In the …rst period, every agent is assigned
a productivity level. Productivity and location are complementary: The more productive
agents are even more productive if they live in the city rather than in the countryside.3
Agents’productivity may change from the …rst to the second period. In particular, there
may be a technological innovation with two e¤ects: an increase both in average productivity
and in the turbulence in the productivity levels of individual agents. Turbulence involves a
reordering of individual productivity levels. For instance, the IT revolution that occurred
in the San Francisco Bay Area in the nineties boosted overall productivity but had a more
positive e¤ect for certain workers (e.g software engineers) than for others (e.g. nurses and
Agents who move to the city in the …rst period have the option to buy or to rent their
house. All houses are identical and can only accommodate one agent. We also assume that
homeownership is a continuous choice variable, going from unboundedly negative (short-
selling city real estate) to unboundedly positive (owning multiple houses in the city, or
derivatives on the city housing price index).4
At the end of the …rst period, city residents vote to determine housing supply in the
second period: They select the number of construction licenses to be issued. A key element
of the model is the institutional mechanism that regulates the distribution of new licenses.
The windfall gain deriving from new construction can accrue to homeowners (e.g. licenses
are sold to developers and the revenues distributed to local homeowners), residents (licenses
are sold to developers and the revenues are used for local services), or to a set of measure
zero of the population (lucky or clever developers in case of an arbitrary mechanism, or
well-connected developers and perhaps corrupt public o¢ cials in case of favoritism). We
2 In Section 6 we will allow for multiple cities.
3 Although we interpret our model in terms of labor productivity, there is a mathematically equivalent
interpretation in terms of quality of life (discussed in detail on page 7). In that interpretation, the bene…t
of living in the city (perhaps a beach resort) is due to local amenities, which are more valuable to certain
people. Within both interpretations, housing is a hedge against rent risk.
4 In Section 6, we study the e¤ect of legal restrictions that make home ownership a binary choice between
renting one home or buying one home.
are particularly interested in situations in which only a small portion of the gain accrues to
residents or homeowners.5
Equilibrium is determined by two orders of interactions. In the housing market, agents
choose how much to invest in real estate. A city resident who does not own housing faces
a rent risk. A positive productivity shock increases the average wage in the city, and hence
rent. The resident can ensure against this risk by purchasing housing. Because productivity
shocks are associated with income turbulence leading to mean reversion, the rent hedging
motive is stronger for agents who currently enjoy high labor productivity. In equilibrium,
investment in housing is therefore an increasing function of the agent’s current productivity
On the political side, agents who are more invested in housing are more likely to favor
a restrictive licensing regime. This leads to a …rst, natural result: In equilibrium, housing
Voters support arti…cial supply restrictions in order to protect their
investment. In turn, they invest because they expect the value of their housing investment
to be protected by urban growth restrictions. This generates an unambiguous welfare loss
because the city remains too small. All citizens, before investing in housing (but after
learning their individual productivity level), would support a commitment to increasing
housing supply to the maximum possible level.
More important, the model displays persistence in housing undersupply. The degree
of undersupply in the second period is an increasing function of undersupply in the …rst
period. This is not due to construction costs (there are none) but to the interaction between
hedging demand and politics. A city with a low initial housing supply is a city with an
initial population of high-productivity agents who pay a high housing price or rent relative
to their income. The median voter is then highly invested in housing, and he is keen to
keep the city small and housing expensive. Under plausible assumptions, there is no new
construction, which leads to the most extreme form of size persistence. The opposite occurs
in a city that starts from a relatively high housing supply and thus low housing costs relative
5 To our knowledge, there is little systematic evidence on the distribution of windfall gains. In her
comprehensive review of housing supply in Britain, Barker (2003, Chapter 5) argues that: (1) Developers
hold option agreements on large tracts of land currently without building permission; (2) Developers have
signi…cant local market power; (3) While local authorities have a legal avenue to demand monetary transfers
in exchange for issuing building permission, the payments obtained in this way are quite low (of the order
of £ 2000/8000 per unit built – See Table 8.2 in Barker). These three facts taken together seem to indicate
that most of the windfall gains accrue to developers. Things may be di¤erent in Hong Kong, where the
government uses an auction mechanism to sell land for development (Wall Street Journal, 10/24/2006).
Once we identify the potentially vicious circle between homeownership and housing
supply, we can begin to discuss the e¤ects of a number of institutional reforms that have
First and foremost, one needs to question the current mechanism for
allocating housing permits. While the allocation system varies widely across countries (and
even within countries), it typically does not take the form of an auction (except for the
Hong Kong example mentioned in footnote 5). Our model formally identi…es a strong link
between the housing undersupply and the share of windfall gains that accrues to the median
voter. The most natural way to break the vicious circle of housing undersupply is to create
simple legal instruments through which local communities can appropriate windfall gains.6
Second, we study the e¤ect of making city planning decisions at a more or less centralized
level. We have assumed that housing supply decisions are taken at a level that corresponds
to the local labor market (i.e. a metropolitan area). In practice, city planning may occur at
a di¤erent level. At one extreme, the U.K. Town and Country Act of 1948 and subsequent
laws give the national government enormous power over planning decisions. The government
can in practice force local communities to accept large-scale land development. At the other
extreme, a number of metropolitan areas around the world (this is true for most large U.S.
cities) are not under a uni…ed jurisdiction: Planning decisions are made by a number of
autonomous local governments. Our paper shows that there exists a U-shaped relation
between the degree of centralization and equilibrium housing supply. A very centralized
system and a very decentralized one result in more construction than a situation in which
local government coincides with local labor markets. A centralized government wants more
housing supply because it takes into account the welfare of countryside residents (who may
move to the city if more houses are built). A very decentralized system falls into a beggar-
thy-neighbor equilibrium whereby local residents do not internalize the negative price e¤ect
that construction in their community imposes on the rest of the metropolitan area.
Third, we examine the e¤ects of subsidizing homeownership. Encouraging households
to own more housing gives them an obvious incentive to restrict urban growth. This is
what happens in equilibrium: When homeownership is subsidized, households vote for a
more restricted housing supply in the second period and housing is more expensive.
6 This point seems to have escaped governments concerned with housing a¤ordability. For instance, the
recent comprehensive report sponsored by the UK Treasury (Barker 2005) uses a wealth of information to
show that housing in‡ation in the United Kingdom is because of an undersupply of land, which in turn is due
to the unwillingness of local authorities to make more land available. However, the policy recommendation
is to tax windfall gains and transfer the proceeds to the central government and to deprive local government
of the only existing channel to appropriate some of the developers’ rent (the so-called “section 106” – see
Barker (2005, p.7, recommendation 29)).
Fourth, we study the e¤ect of imposing restrictions on fractional ownership. Caplin et
al. (1997) argue that current rules make it di¢ cult for people to share ownership of their
home with others. We compare the baseline case with a set-up in which people can only
own zero or one home. Besides the direct portfolio e¤ect discussed by Caplin et al., we
identify an indirect supply e¤ect of restrictions to fractional ownership. The elimination
of such restrictions is likely to make housing more a¤ordable through increased political
support for urban growth.
Our paper brings together two in‡uential strands of literature on housing that have
hitherto remained mostly separate.
The …rst strand moves from the premise that a meaningful discussion of the housing
market must include an endogenous housing supply function. Glaeser, Gyourko and Saks
(2005b) provide a detailed model of the decision process involved in authorizing housing
development. They study the e¤ects of changing judicial tastes, decreasing ability to bribe
regulators, rising incomes and demand for public amenities, and improvements in the ability
of homeowners to organize and in‡uence local decisions. They …nd that a signi…cant increase
in the ability of local residents to block new projects is the main driver for the rise in urban
growth restrictions. They conclude that cities have changed from urban growth machines
to homeowners’cooperatives. Our paper forgoes most of the political process complexity of
Glaeser et al. in order to endogenize the composition of the local population, households’
tenure decision, and hence their preferences for urban growth.
Our approach to the voting decision builds on the work of Fischel (2001), who provides
detailed arguments and empirical evidence that downside risk motivates homeowners to
participate in the planning process. He argues that “homevoters” are politically motivated
by the risk of loss on their home because of the di¢ culty to diversify this risk away. House-
holds in our model are motivated not only by potential loss because of new construction
but also by the prospect of capital gains when aggregate demand increases.
The second strand of literature endogenizes housing investment by modeling the tenure
choice of risk averse households in a stochastic environment (e.g., Ortalo-Magné and Rady,
2002, Sinai and Souleles, 2005, Hilber, 2005, Davido¤, 2006, Shore and Sinai, 2006). This
strand of literature links housing prices with expected future rents and shows that home-
ownership provides a hedge against future housing expenditures. Sinai and Souleles (2005)
developed a stylized model of dynamic investment decisions by households facing stochastic
rent ‡uctuations and endogenous house prices, and they show that it can account for a
number of observed patterns in tenure decisions.
Our model incorporates this key insight about housing demand in a market equilibrium
model. The rent risk is now endogenous, and it is the combined e¤ect of labor productivity
shocks, households’location decisions, and collective supply decisions. As in the contribu-
tions cited above, the possibility of rent ‡uctuations gives housing a hedging value, which
pins down homeownership patterns and, in turn, determines political support for urban
While our model brings together these two frameworks, it neglects – by necessity – a
number of important issues related to housing supply. In particular, we abstract from the
issue of local taxation for the provision of local public goods, peer e¤ects and agglomeration
As several papers have shown, externalities are essential to understanding
the political economy of housing supply. In this paper, we abstract from them in order
to clarify the dynamic connection between homeownership and housing supply. We will
mention where appropriate how our model could incorporate local externalities. Also, our
model considers only one – stylized – form of growth restriction. We do not consider
regulation pertaining to height, density, use, etc.9 This choice, again, is in the interest of
parsimony. Our methodology can easily be extended to other forms of regulation.
The plan of the paper is as follows. Section 2 lays out the model. For expositional
purposes, we …rst analyze the model holding housing supply …xed (Section 3) and we then
endogenize supply and study the political equilibrium (Section 4). We discuss the persis-
tence result in Section 5. Section 6 studies the e¤ect of institutional reforms. Section 7
concludes. All proofs are in the Appendix.
7 One may wonder whether the link between preferences over urban growth and individual labor produc-
tivity levels could be obtained, in a deterministic model, through a simple wealth e¤ect. However, the fact
that rich people consume more housing than poor people is not enough. One must also argue that rich
people have a reason to spend proportionally more on purchases rather than on rentals, which – in a world
without uncertainty –requires ad-hoc assumptions (e.g. rental-related transaction costs are relatively higher
than purchase-related transaction costs for expensive properties or the tax advantage to owning is increasing
8 See Fernandez and Rogerson (1997) and Calabrese, Epple and Romano (2006) for studies of the political
economy of zoning regulation and its interaction with the provision of local public goods.
A number of papers analyze growth controls in a static setting. They focus on issues related to the
creation of amenities and the distribution of the value generated by the location of economic activity within
cities; e.g., Brueckner (1995), and Helsley and Strange (1995), Brueckner and Lai (1996). As mentioned in
Brueckner and Lai, a dynamic model is necessary to capture the motivation for growth controls that comes
from the prospects of capital gains on one’s home.
9 For references, see for example Wheaton (1998), and Bertaud and Brueckner (2004). Note that Glaeser
and Ward (2006) …nd that a variety of regulations, which act as very e¤ective barriers against new construc-
tion, generate little price e¤ects beyond their e¤ect on housing density. For example, if high minimum lot
sizes were used to “improve” the mix of a community and such “improvements” had a signi…cant economic
impact, their housing price e¤ect would be greater than that justi…ed by their restrictive e¤ect on the number
of homes that can be built
Consider a two-period model of an open economy with two locations, the city and the
countryside. There are two commodities, housing and a numeraire consumption good.
For simplicity, we abstract from housing construction costs. Housing in both communities
consists in homogeneous plots of land. The economy is populated by a mass 1 of agents
with identical CARA utility de…ned over consumption of second period numeraire only,
The endowment of numeraire good each agent receives every period depends on his loca-
tion and his productivity index i 2 [0;1]. We normalize the productivity in the countryside
to zero.10 Let yi be a random variable uniformly distributed on [0; 1]. In the …rst period,
if agent i works in the city, his productivity is yi1 = yi. In period 2, the distribution of
earnings evolves as follows. With probability 1
, no shock occurs: if agent i works in
the city, his productivity is yi2 = yi1 = yi. With probability , there is a positive shock to
aggregate productivity: average productivity increases by g > 0. Conditional on this shock
occurring, with probability 1
; the productivity of agent i in the city is yi2 = yi + g. With
probability , all agents draw a new productivity parameters ~
yi from the initial productiv-
ity distribution. The productivity of agent i in the city is then yi2 = ~
yi + g. The expected
aggregate growth rate of the economy between periods 1 and 2 is therefore g. The greater
the probability , the more insecure are the agents about their future productivity.
Our key assumption about the labor market is that there is positive correlation between
growth and turbulence. If the economy grows faster, there is a higher probability that the
income ranking in period 1 is changed in period 2. Such a positive correlation could be
because of technological change. Every time an important innovation is introduced, most
agents reach higher productivity levels but certain agents, whose skills become obsolete,
lose out, at least in relative terms.11
This same mathematical formulation also has a di¤erent interpretation in terms of leisure
rather than productivity. The variable yi represents the utility that agent i gets if he lives
in the “city”, which can now be seen as an area with certain amenities. Such amenities
may be natural (a coastal region, a ski resort) or man-made (a historical town, a vibrant
1 0 Our results could easily be extended to a more general setting in which the productivity of agent i at
time t is yit in the city and ayit in the countryside, where a 2 (0; 1).
1 1 An alternative model, which would yield qualitatively similar results, is one in which a positive shock
corresponds to random waves of new groups of skilled professionals. Such arrivals produce a twofold e¤ect:
The city reaches a higher average income, but the previous city residents are now relatively poorer that the
metropolis). Shocks in yi derive from social phenomena that determine shifts in preferences
for amenities (the desire to retire to sunnier climates, the feasibility of telecommuting,
reduced crime rates in large cities). Such shifts create both a higher average utility of living
in the “city”and turbulence in the utility rankings of agents. For concreteness, the rest of the
paper will refer to the productivity interpretation rather than to the leisure interpretation.
Working in the city requires consuming one unit of city housing. We denote lit the
housing consumption by agent i in period t where lit = 1 if the agent locates in the city,
lit = 0 otherwise.
Independently of their housing consumption choice, agents may also invest in city hous-
ing. Let hit 2 ( 1;1) be the measure of city housing that agent i owns in period t. We
do not restrict this measure to be a positive integer. A noninteger hit indicates fractional
property. A negative hit means that the agent has sold city housing short for period t. In
practice, there are serious obstacles to fractional property and to shortselling properties.12
In the countryside, the supply of housing is perfectly elastic at a cost normalized to
zero. There are no moving costs between city and countryside.
There is a measure N1 of housing in the city at the start of period 1 owned initially by a
a large number of international real estate investment trusts (REITs), which maximize the
expected value of their real estate investment. At the end of period 1, city residents choose
the measure N2 of houses available in the city in period 2. We assume existing houses
cannot be destroyed and no depreciation of the housing stock, N2
N1. city residents
therefore vote on the number of building permits that will be issued, N2
Permits are assumed to be identical, divisible, and immediately tradeable. Given the
2 [0;1] and 2 [0;1], the permits are allocated as follows:
A proportion 1
of the permits goes to a set of measure zero of the population. We
think of this as the classical system of allocating permits to certain developers without
asking them for a payment corresponding to the capital gain they will experience.
These developers may in turn return some of the windfall gain to city o¢ cials in the
form of bribes or campaign contributions. We assume that developers and o¢ cials
represent a negligible proportion of the city population (and hence the median voter
never bene…ts from this share of permits).13
1 2 We thus abstract for the time being from a number of potential imperfections of the housing mar-
ket (later, we will discuss the e¤ect of restrictions to fractional properties and the introduction of …scal
1 3 The writer Tom Wolfe (New York Times, 2006) o¤ers a vivid account of the permit allocation process
in New York City. A key role is played by the Landmarks Preservation Commission, a body that decides