The Economic Journal, 110 (October), 939±962. # Royal Economic Society 2000.

Published by Blackwell
Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA.

THE ECONOMICS OF TENANCY RENT CONTROL

Kaushik Basu and Patrick M. Emerson

We consider a rent control regime where rent increases on, and eviction of, a sitting tenant are
forbidden. When apartments become vacant landlords may negotiate new rents. If in¯ation
exists, landlords prefer to rent to short-staying tenants. Since departure-date-contingent
contracts are forbidden and landlords cannot tell whether tenants are short-stayers, an adverse
selection problem arises, with a Pareto inef®cient equilibrium. When tenant types are
determined endogenously, multiple equilibria can arise where one equilibrium is Pareto
dominated. Abolition of the rent control regime, cannot only shift the equilibrium out of this
inferior outcome, but also result in across-the-board lowering of rents.

In early 1996, when New York City's rent control law came up for evaluation
and possible modi®cation, the public debate spilled over beyond New York to
national newspapers and the international media. The same questions that
arose in this debate have arisen in the past in discussions concerning rent
control in, among other places, France, Germany, India, Sweden and other
parts of the United States. These debates reveal, more than anything else, how
widely the central issues of rent control are misunderstood. Part of the blame
for the popular misunderstanding of the effects of rent control lies with
economists. Despite quite a substantial literature on the subject, some of the
key analytical questions, especially ones concerning the relation between
in¯ation and rental adjustment, remain unanswered.
The aim of the present paper is to construct a model that captures the main
stylised features of a form of rent control pervasive around the world. We refer
to it as `tenancy rent control'. Tenancy rent control, which is a special case of
what is known in the literature as `second-generation rent control', allows
landlords to choose a nominal rent freely when taking on a new tenant (the
tenant is of course free to reject the offer) but places restriction on raising
rents on, or evicting, a sitting tenant. This causes an erosion in the real value
of rent if a tenant stays on for too long, whenever there is positive in¯ation in
the economy, which for most economies is true most of the time. This means
that landlords will prefer short-staying tenants to long-staying tenants. Since a
tenant's type will be better known to the tenant than the landlord, the tenancy
market will be characterised by asymmetric information. Our basic model
describes the tenancy market as a model of asymmetric information in which
the tenants' types are exogenously given. It is shown that the presence of
tenancy rent control will, in general, result in a Pareto sub-optimal equili-
brium, whereas a system of free contract will be Pareto optimal. Of course, this

 The authors have bene®ted from presentations of this paper at the American Real Estate and
Urban Economics Association Annual Conference and at the Applied Microeconomics Workshop at
Cornell University. For comments and discussion, we would like to thank Richard Arnott, Pinaki Bose,
Franz Hubert, Robert Masson, Ted O'Donoghue, Edgar Olsen, Buhong Zheng and two anonymous
referees of this Journal.

[ 939 ]
940 THE ECONOMIC JOURNAL [OCTOBER
does not mean that moving from the former to the latter would make every-
body better off. However, the model does illustrate how the real con¯ict of
interest is not between landlords and tenants, as portrayed in most popular
debates on rent control, but between tenants of different kinds. This is a result
that will come as no surprise to economists. In our case this basic model and
result serve as a benchmark that can thereafter be developed to obtain some
surprising results.
After constructing the basic model we develop it by endogenising the tenant
types. That is, we allow for the fact that the outcome in the rental market may
affect the tenant's life-style, for instance, discouraging him from shifting too
many times. Once the tenant's `type' is modelled as an endogenous variable
we get the surprising result that rent control may give rise to multiple
equilibria. This is a very natural result, based on weak and realistic assump-
tions, but it seems to have gone unnoticed in the literature. If the economy
gets locked in the `bad equilibrium', among the many possible equilibria, the
removal of rent control cannot only bring about an ef®cient outcome but
cause an across-the-board lowering of rents, thereby leaving all tenants better
off. This result is established in Section 4.
We should clarify that all our comparisons of different rent-control regimes
take the form of comparative statics. We do not consider switch-overs from one
regime to another. Hence all of the policy prescriptions that ¯ow out of this
exercise concern new tenants and new contracts. We do not comment on how,
or for that matter whether, changes should be made to laws applicable to
currently sitting tenants.
The next Section is about the institution of rent control. It discusses differ-
ent kinds of rent control, some stylised facts, and the real-world context of our
theoretical constructions.

1. The Institution of Rent Control

In the United States and Europe, the numerous governmental controls in the
rental housing market, which are generally described as `rent control', arose
during World War II in response to the mass disruptions caused by the war.
After the war, New York City and many European jurisdictions retained versions
of these out of fear that the return of troops would send rents skyward. In other
parts of the United States, the social upheaval and high in¯ation of the 1970s
was a driving force behind the re-implementation of rent controls. California,
Connecticut, Massachusetts, New Jersey and New York are all states where
jurisdictions implemented rent control policies during this period. While a
number of jurisdictions have since abandoned or relaxed rent control laws,
these laws are still commonly found in the United States and the world over.1
Before we proceed to discuss rent control, it is worthwhile clarifying that the
absence of rent control can be of two kinds: the kind with no government

1
Arnott (1995) discusses the history of rent control in the United States and Europe and provides a
useful bibliography.
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 941
intervention in the rental housing market, or the kind where the government
allows and enforces contracts (subject to the standard restrictions on the
freedom of contract provided under contract law). In this paper, when we
consider a regime with no rent control, we shall be concerned with the latter,
which will be referred to as a `free contract' regime. By its converse, `rent
control' is a generic term that describes rental laws, which place additional
restrictions on allowable contracts.
While many different forms of rent control exist in the world, we will focus
on a stylised version, which is widely used throughout the world. We will focus
on a rent control regime that does not allow the eviction of a sitting tenant
and that limits the amount a landlord may increase the rent on a sitting tenant
(enough so that rents do not, typically, keep up with in¯ation). Upon vacancy,
however, the landlord is free to negotiate a new rent with a new tenant. This is
precisely the regime that exists in quite a few US communities, including Los
Angeles, Berkeley, Santa Monica and Palm Springs and is similar to the system
that exists in Washington, D.C. (Dreier, 1997, and Olsen, 1990).2 This is also a
good approximation of rent control laws in other communities in the United
States and elsewhere in the world including Germany, France (Hubert, 1995,
and BoÈrsch-Supan, 1986), Sweden, and virtually all major cities in India.
In Delhi, for example, Section 6 of the Delhi Rent Control Act, 1958, allowed a
maximum of a 10% rent hike every three years, no matter what the in¯ation. In
India the average in¯ation every three years has exceeded 20%. The Act also
made it virtually impossible to evict a tenant. The 1958 Act has subsequently
been superseded by the Delhi Rent Act, 1995, which is only slightly more ¯exible.
In New York City, properties under `rent stabilisation' are closest to the rent
control law just described. There have been two major rent control regimes in
New York, `Rent Control' and `Rent Stabilisation.' Rent Control was a strict
regime started in 1947 that assigned rents for individual properties and
allowed minor increases. This policy currently covers slightly more than 70,000
units in New York City and is declining with vacancy decontrol and shifts to
stabilisation. Much more common are properties under Rent Stabilisation.
This system was implemented much later, in 1969, and was a less stringent
form of rent regulation where periodic rent increases are allowed as approved
by a rent regulatory board. In 1971, a policy of vacancy decontrol was instituted
for all units under either Rent Control or Rent Stabilisation after they had
been `voluntarily' vacated. However, in 1974 an amended policy was intro-
duced that ended vacancy decontrol of stabilised units and controlled units in
buildings with more than ®ve units in total. The new policy kept the units
under stabilisation, but allowed increases of between 14 and 16% above the
most recently approved allowable rent increase for a voluntarily vacated
apartment.3 Recently, the 1997 agreement to renew New York's Rent Stabilisa-
tion laws upped the allowable vacancy increase to 20%, for a two-year lease,

2
In Los Angeles annual rent increases for sitting tenants are limited to 7%, but once vacated, a new
rent can be freely chosen.
3
See Linneman (1985) for a good discussion of the history of New York City rent control.

# Royal Economic Society 2000

942 THE ECONOMIC JOURNAL [OCTOBER
and an additional 0.6% for each year the previous tenant had occupied the
unit.4 There are other ways for landlords to raise the controlled rent over and
above these increases as well. One way is through pass-through costs. This
allows landlords who spent money on improvements to a rent stabilised
apartment to raise the rent by appealing to the board. Another way is by
pleading hardship or increased operating costs. Anecdotal evidence suggests
that landlords raise rents through these channels often, even if the amount of
money spent on improvements is small. As both of these tactics are sure to
meet opposition from a sitting tenant, the common practice is for the landlord
to appeal for these increases upon vacancy.5
New York City's rent control laws also provided non-rent protection for
tenants. In particular, the Rent Stabilisation Code stipulated that landlords
must offer tenants a renewal lease (at the stabilised rent) before the expiration
of the current lease. It also limited the set of circumstances in which the
landlord could evict a sitting tenant (non-payment of rent, for example).6
These provisions are essentially the non-eviction of a sitting tenant clause that
we stipulate for the stylised rent control law studied in this paper. Thus, the
current situation in New York City is very close to the tenancy rent control
regime of our model. Therefore, it seems to us, that tenancy rent control is
pervasive, especially when one recognises that a rent control system applied to
a unit rather than the tenancy, in which rent increases are permitted unless
appealed against by the tenant, is not very different.
There has been a considerable amount of theoretical work on rent control.
The textbook version of rent control is a price-ceiling model of supply and
demand that relates most closely to what Arnott (1995) describes as `®rst
generation rent control'. These are akin to the rent control regime, which
New York City implemented in 1947, in which rents were ®xed at a level and
rarely allowed to rise. The textbook model has been advanced in several
directions (see, for instance, Hubert (1996); Raymon (1983); Frankena
(1975); Sweeney (1974); for a related survey, see Smith et al. (1988)). There
has also been work done on the political economy of rent control (see Epple
(1998) and Fallis (1988)), which try to explain why the rental laws are what
they are by looking at the power structure of different lobbies.
Our aim is however more limitedÐto study the economics of tenancy rent
control in economies with positive in¯ation.

2. The Basic Model

Let us assume that there are n types of potential tenants in an economy. If N
is the set of types, then N ˆ f1,


, ng. Suppose a fraction pi of all tenants are
of type i. Thus p1 ‡


‡ pn ˆ 1. All agents in this paper are in®nitely-lived.

4
`Deal Is Achieved as Rent Laws Expire' by James Dao, The New York Times, June 16, 1997.
5
See Jarett and McKee (1997) for anecdotal evidence of the rent increasing tactics of NYC landlords
as well as a brief history of rent control in New York City.
6
Cinque (1997) discusses the non-rent protections afforded tenants by New York City's various rent
control laws.
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 943
A tenant's type basically refers to how long a tenant stays in the same
apartment before moving to a new one. Let t i be the number of months a
tenant of type i stays in the same apartment. Without loss of generality we
assume that,

t1 , t2 ,


, tn:

In other words, type 1 tenants are the restless souls. Either they have a
preference for quick change or have transferable jobs. Type n tenants are the
types who gather moss. Others are somewhere in between those extremes. Of
course, in reality, depending on the rent-control regime that prevails in an
economy, a person may decide to quit a transferable job and take up a stable
job or vice-versa. But we will, for now, assume that the tenant types are given.
This assumption is relaxed in the next section.
Throughout this paper we assume that there is a ®nite number of tenants
and a ®nite (and therefore discrete) number of tenant types; and that rents
are paid at discrete time intervals. But our method and all the results in this
Section extend easily to the case where there is a continuum of tenant types
and rents are paid continuously.
This is a model with asymmetric information. Each tenant knows his type
but a landlord cannot tell the tenant's type by looking at him. In addition, our
rent-control law does not allow quit-contingent contracts, rent escalation
clauses for long-stayers, nor length contingent payments to tenants. The
monthly rent has to be ®xed at the time of taking on a new tenant and may not
be adjusted for the duration of the tenancy.
Note that even though a tenant's type is unknown to the landlord at the
time the tenant moves in, the tenant's type gets revealed at the time the tenant
moves out. Hence, by charging a lump-sum amount at the time of a tenant's
moving out, a landlord can overcome the problem of asymmetric information.
A rent-control law typically prevents such complicated contracts and thus
causes the asymmetric information problem to persist (Basu, 1989), in fact
without such restrictions a rent control law would be rendered ineffectual.
Initial deposits such as key money (or what in India is called pugree) with
agreement to return a part of it depending on when the tenant leaves, or any
kind of rent escalation clause may be viewed as the market's way to get around
rent-control. In our model we assume that these types of payments and clauses
are not allowed by the law. In other words we are about to analyse the case of
`tenancy rent control,' as described in Section 1.
We will also assume that there is in¯ation in this economy which erodes the
value of money each month by 1 ÿ â, which is greater than zero. That is, we
are assumingÐwhat is nearly universalÐthat there is some in¯ation in the
economy.
Let the discount factor for all individuals be ä 2 (0, 1), for each month.
If a landlord charges a rent of 1 dollar per month in real terms from a new
tenant and somehow gets only tenants of type i, then the stream of income
earned by the landlord, in real terms, is given by
# Royal Economic Society 2000
944 THE ECONOMIC JOURNAL [OCTOBER
1 â â2


â tÿ1 1 â â2


â tÿ1 1

Given the presence of rent control, this stream of income is easy to under-
stand. Since the market rent for a new tenant is 1, the landlord earns 1 in
period 1. Since the rent control law does not allow the nominal rent to be
changed, and the in¯ation rate is 1 ÿ â, in the second period (third period)
the 1 dollar is equal to â dollars (â2 dollars) in real terms. This explains the
second and third terms in the stream shown above. After t periods the tenant
quits. The new tenant pays a rent of 1 dollar in real terms (or âÿ t dollars in
nominal terms). This explains why the t th term is 1 and so on. The present
value of the above stream using the discount factor of ä is denoted by v i and
this is given by:

v i ˆ 1 ‡ âä ‡ ( âä)2 ‡


‡ ( âä) t i ÿ1 ‡ ä t i v i (1)

or,

1 ÿ ( âä) t i
vi ˆ : (2)
(1 ÿ âä)(1 ÿ ä t i )

Lemma 1. If i , j, then v i . v j .
The proofs of this and of all following lemmas (save lemma 3 for which an
intuitive proof is given in the text) are found in Appendix A.
Continuing with the case in which rent is $1 per month, let us denote v(i) as
the expected present value of returns to the landlord when all tenants of type i
or above make themselves available to the landlord as potential tenant from
whom the landlord randomly selects one.7 Then, clearly v( n) ˆ v n . And, more
generally,
0 1
P
n pk 2 t k ÿ1
v(i) ˆ BPn C[1 ‡ âä ‡ ( âä) ‡


‡ ( âä) ‡ ä t k v(i) ] (3)
kˆi @ A
pj
jˆi

or

7
In our framework we are taking this environment to be static and thus we present a static model,
however it is important to note that in the dynamics of the model short stayers will appear more
frequently on the market for rental housing than will long stayers. What this means, in effect, is that the
present value of the returns to the landlord when facing a mix of indistinguishable potential tenants
should not depend on the proportion of types in the economy (as in (3)) but rather the proportions of
each type that are in the market for a vacant apartment at the steady-state. While our method is a
simpli®cation, using the steady-state proportions will yield the same asymmetric information results as
we derive using our method, however a dynamic model of this situation would be a worthwhile
undertaking in the future.
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 945
0 1
P
n pk 2 t kÿ1
BPn C[1 ‡ âä ‡ ( âä) ‡


‡ ( âä) ]
kˆi @
p A j
jˆi
v(i) ˆ 0 1 : (4)
P
n pk tk
1 ÿ BP n Cä
kˆi @ A
pj
jˆi

Since the above expressions are worked out assuming that the rent is one
dollar (in real terms for a new tenant), it is now easy to work out the
expressions for the case when the rent is R dollars. If the landlord gets only
tenants of type i, we denote the present value of her rental income as v~ i (R)
and clearly
~ i (R) ˆ Rv i
v (5)
where v i is given by (2).
If the rent is R and only tenants of type i and higher seek tenancy, we
denote the landlord's present value of income form leasing out one apartment
~(i) (R). Clearly,
by v
~(i) (R) ˆ Rv(i)
v (6)
where v(i) is given by (4).

Lemma 2. If i , j, then v(i) . v( j) .

As we discussed in the introduction, one of the most popular variants of rent
control takes the form of disallowing landlords from raising rents adequately
or evicting tenants. Let us, in particular, assume that a landlord can choose a
rent, R, at the time of taking on a new tenant (who of course has the freedom
to turn down the offer); but then the rent remains (nominally) the same as
long as the tenant stays on.
Let us now model the tenant's decision-problem under this regime of
tenancy rent control. Let us assume that all tenants have the same option
(irrespective of their types) if they reject leasing an apartment.8 They could
locate in an area not covered by the rent control law (most often suburban
communities, but, as in the case of New York City and Los Angeles, sometimes
rent control coverage is only partial and rent controlled apartments exist side-
by-side with non-rent controlled apartments), buy a house, live off friends or
live in motel rooms or mobile homes, which are not covered by the rent
control ordinances. That the preferred outside option is the same across types
is reasonable if one imagines that, for example, the suburban rental market is
competitive. Thus rents will be set to cover costs and will be freely adjusted to
in¯ation, neither of which depend on tenant type. Of course living in a suburb
often involves additional transportation costs to those who work in the city

8
As explained later, much of our results would continue to hold without this assumption; but it is a
useful simplifying assumption and will be used throughout.
# Royal Economic Society 2000
946 THE ECONOMIC JOURNAL [OCTOBER
making the option of locating to them less desirable all else equal. In a
competitive environment, however, if the cost of making an apartment avail-
able to rent is the same across landlords (i.e. in both rent controlled and non-
rent controlled areas) then even if renting in the non-controlled areas involves
no additional costs to tenants, the results of the model will hold. The landlords
zero pro®t condition assures this.
The outside option gives a person a life-time utility of B. We assume that all
tenants receive the same life-time bene®t from renting an apartment, A, and
must, of course pay rent R, which, in present values terms is Rv i for a type i
tenant. We assume that A . B, and de®ne the difference, A ÿ B, as D. There-
fore a tenant will lease an apartment if and only if, A ÿ Rv i > B or
Rv i < A ÿ B  D. What we mean by this, in operational terms, is: Irrespective
of a tenant's type (which is here exogenously given), if a tenant ®nds that the
present value of rentals exceed, D, the tenant will opt out of tenancy.
If a tenant is of type i, and the rent is R, the present value of rentals paid by
the tenant is clearly Rv i , as in (5), with v i as de®ned by (1) or (2). Hence, a
type i tenant will opt for tenancy as long as Rv i < D. By Lemma 1 we know
that as R increases, the shortest-staying tenants (i.e. of type 1) will be the ®rst
ones to opt out of tenancy, followed by types 2, 3 and so on with the last to opt
out being the longest stayers (type n). Since the short stayers are the more
attractive tenants from the landlord's point of view, this is what drives the
adverse-selection process in this model.
Now consider a landlord who has one property to lease out. Let V (R) be the
landlord's expected present value of rental when the per-period rent is R.
Following the argument in the above paragraph, we can now compute what
V (R) will be like as R varies. An important and interesting implication of this
is the following.
Lemma 3. V (R) reaches its maximum when R ˆ D=v n .
Note that D=v n is the critical rent above which the longest-staying tenants
opt out of tenancy. The proof of Lemma 3 is obvious with the use of a
somewhat unusual diagrammatic technique that we develop below. Let us ®rst
explain how V (R) can be represented diagrammatically. Consider a case
where n ˆ 3. In Fig. 1, the horizontal axis represents, R. In this Fig. draw the
lines Rv1 , Rv2 and Rv3 . By Lemma 1, Rv1 is the steepest followed by Rv2 and
then Rv n or (what is the same here) Rv3 . Draw a horizontal line at height D
from the origin and mark off the critical rents, namely, D=v1 , D=v2 and D=v3
where each type drops out of the rental market. All this is shown in Fig. 1.
Being a model of asymmetric information it is not surprising that the ®gure is
similar to the one in Stiglitz and Weiss (1981) which plots bank return as a
function of the interest rate.
In the same ®gure draw Rv(1) , Rv(2) and Rv(3) . Recall that Rv(i) is a
weighted average of Rv i , Rv i‡1 , . . . and, Rv n . It follows that Rv(3) coincides
with Rv3 .
Now suppose the monthly rent is below D=v1 . Then all three types of tenants
seek tenancy. Hence the landlord's expected present value of rentals earned
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 947
Rv1 Rv2 Rv(1) Rv(2)

Rv(3)5Rv3

V(R)

D/v1 B D/v2 D/v3 Monthly

Rent, R

Fig. 1. The V(R) Function

(i.e. V(R)) is given by Rv(1) . Once R exceeds D=v1 , type 1 ceases to lease in
property. As long as R < D=v2 , V (R) is equal to Rv(2) . Beyond D=v2 , V (R)
equals Rv(3) . Hence the landlord's expected present value of rentals, V(R),
must satisfy the following:
8
>
> Rv(1) , if R < D=v1
<
Rv(2) , if D=v1 , R < D=v2
V (R) ˆ
>
> Rv , if D=v2 , R < D=v3
: (3)
0, if D=v3 , R

The V (R) function is illustrated by the thickened line in Fig. 1.

One easy implication of the de®nitions is that Rv(i) , Rv i for all i , n. It
immediately follows that all the peaks of V (R) excepting the one at rental,
D=v n , will be dominated by the peak at D=v n (as shown in Fig. 1). This
completes the proof of Lemma 3.
Note that if the opportunity cost of leasing property differed for each type,
that is, the opportunity cost to type i was Di instead of D, Fig. 1 would have to
be adjusted by drawing in D1 , D2 , and D3 and locating the critical rentals
D1 =v1 , D2 =v2 , and D3 =v3 . Lemma 3 would not then be necessarily valid.
However, the inef®ciency results that we prove do not hinge critically on the
assumption of `equal opportunity cost.' And so, in view of its simplifying
nature, this is an assumption that we continue to use.
# Royal Economic Society 2000
948 THE ECONOMIC JOURNAL [OCTOBER
It is now easy to see the that if the rental market is monopolistic (in the
sense of there being one landlord with 1 property to lease out) then she would
set the rent at D=v n . All but the longest-stayers would be driven out of the
market in equilibrium. Consider the extreme case where t n ˆ 1. In this case
rental is equivalent to sale by instalment payment, where the instalment
payments go on forever.
Let us now proceed to analyse what happens if there are many landlords
competing with one another as would be the case in any large city. Let us
assume that the cost to a landlord of leasing out an apartment (in present-
value terms) is C. In order to consider the interesting case suppose C , D. It
follows therefore that it is Pareto ef®cient to let all tenants have a rented
property each. But let us see what the outcome will be under perfect competi-
tion. By perfect competition we mean here that (a) all agents are price-takers
(which implies, in particular, that if the market rent is R, a landlord expects to
get no tenant if she unilaterally raises the rent to R9) and (b) there are enough
(potential) landlords to drive pro®t down to zero.
The perfectly competitive outcome is easy to illustrate using the diagram-
matic technique developed above. In Fig. 2 we reproduce the V (R) curve. Let
us suppose that C is as shown, and R  be such that V (R  ) ˆ C.
If the market rent, R, exceeds R  , the landlords will be making supernormal

V(R)

C′

D/v1 D/v2 R′′ R∗ D/v3 Monthly

R′ Rent, R

Fig. 2. Rent Control Equilibrium

# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 949
pro®ts. So there will be more entry of landlords and R cannot be an
equilibrium. If R is below R  , C . V (R) and landlords will exit the rental
market. Hence R  (in Fig. 2) is the equilibrium rent under perfect competi-
tion when landlords cost of leasing out property is given by C.
Let us now consider the case where the landlord's cost, instead of being C, is
C9. Then there are two rental rates, R9 and R 0 at which C9 ˆ V (R9) ˆ V (R 0).
Are both of these equilibria? It depends on how exactly we interpret a
`competitive market.' If by that we mean that landlords cannot change rents in
either direction (as in Mas-Colell et al. (1995), Chapter 13) then both R9 and
R 0 constitute equilibria. However, it seems reasonable to argue that while no
landlord can unilaterally deviate from the market rent in an upward direction,
they can unilaterally deviate in a downward direction (without losing all
tenants). Then R 0 ceases to be an equilibrium. First, we explain this intuitively;
and then (at the end of this section) we give it a formal game-theoretic
interpretation.
The intuitive reason is simple. Suppose all landlords are charging R 0. Then
if one landlord cut her rent to R^ 2 (R9, D=v1 ), all tenants would try to lease
from this landlord and she would make a positive pro®t, since (as is clear from
^ . C9. Hence, R9 is the only competitive equilibrium in this case.
Fig. 2) V ( R)
Let us see how different people are affected in equilibrium. By comparing
Figs 1 and 2 it is clear that the cost of rental to types 1, 2 and 3 are R9v1 , R9v2
and R9v3 . By marking the R9 point of Fig. 2 in Fig. 1 it is clear that
R9v1 . R9v2 . R9v3 and R9v2 . C9 . R9v3 . Hence the short stayers pay rents
that are too high (above the cost to the landlords) and long stayers pay rents
that are too low (type 1's do not rent at all in this market as R9v1 . D). This is
the real dividing line in the rent control debate.
Most of the popular divisions arise between landlords and tenants. If the law
is going to be changed on sitting tenants, indeed there would be con¯icts of
interest between landlords and tenants. But if the law is going to be changed
for all future tenancy contracts (as it should be) then the con¯ict of interest is
between tenants of one type and another. But these are such diffuse categories
that popular attention has mistakenly been directed at the more visible line
between landlords and tenants.
Another popular misconception is the indignation people feel when they
hear of people who have held apartments for many years and now pay
`absurdly' low rents ± the old widow in Delhi or New York who pays 100
Rupees or Dollars per month for her two-room apartment. This indignation
has, however, nothing to do with our model but, in fact, arises from a
misunderstanding of the economics of rent control. Suppose the old widow
had bought the apartment many years ago (for a price which today would
appear absurdly small). This is equivalent to a rental agreement which involves
a lump-sum initial payment and then a very small monthly payment, in this
case, zero. Should this be reason for shock and indignation? Should this be the
basis for saying that the old lady should return the apartment to the original
owner or at least start paying a higher monthly rent? Clearly not. She bought
the place and that is the end of the matter. But now, if a monthly rent of zero
# Royal Economic Society 2000
950 THE ECONOMIC JOURNAL [OCTOBER
is no reason for indignation, why should a monthly rent of Rs. 100 be treated
differently? If the old lady pays a rent of Rs. 100 because the agreement (or
generally accepted presumption) was that that is what she would do forever,
then we could, effectively, think of her as having bought the place under the
agreement to pay in monthly instalments of Rs. 100.9 Of course, Rs. 100 looks
very small today; but zero looks even smaller and that is what a person who
bought the apartment she lives in pays. Indeed in a free contract system,
someone may sign exactly the contract the old widow signed and thirty years
down the road the rent will appear `absurdly' low. This, therefore, is clearly
not the essential difference between a system of rent control and one of free
contract.
Return to the case illustrated in Fig. 2 where the landlord's cost is C. In
equilibrium everybody excepting the type-n tenants are driven off the market.
Yet for each tenant-type, i, there exists a rental such that both landlord and
tenant would bene®t. This implies that under rent control the competitive
equilibrium can be Pareto inef®cient.
If the rent control law is revoked and replaced by a system in which a tenant
and a landlord can sign on any rental and eviction contract, it is easy to see
that Pareto ef®ciency is attained. This is because the asymmetric information
problem is not insurmountable here. Though for a new-tenant his type is not
transparent to the landlord, at his time of departure his type gets revealed. So
by writing a departure-date contingent contract (or by putting in a rent-
escalation clause) landlords can get around the asymmetric information
problem. The problem with most rent-control regulations is that they tend to
disallow (or render illegal) clauses in rental contracts that allow landlords to
overcome the asymmetric information problem.
It is also worth noting that landlords may try to counter some of the
disadvantages of tenancy rent control by using harassment. If a landlord
expects to bene®t from a tenant vacating his unit, he will have the incentive to
harass the tenant and inadequately maintain the property. This could explain
why certain kinds of deposit money are discouraged under the law.
Let us, for completeness, consider the case where a landlord and a tenant
can agree to any contract and the state legal machinery ensures that the
contract is adhered to. Under such a legal regime, one kind of contract that
would achieve optimality is a fully-in¯ation indexed rental contract. Suppose a
landlord writes a contract where the real rental is R each month. In other
words, the nominal rent is raised each month suf®ciently to correct for the
amount of in¯ation. Under such a rental contract, the tenant type is unim-
portant to the landlord because no matter who the tenant is, if the real
monthly rental is R, the landlord's present value of earnings is R=(1 ÿ ä). If
the cost of leasing out is C, the competitive real rental rate is R  where
R  =(1 ÿ ä) ˆ C.

9
There may be inef®ciencies arising in this case because the original contract was unclear or because
a tenant has limited rights compared to an owner (Basu, 1989) but those are not our concern in this
paper.

# Royal Economic Society 2000

2000] THE ECONOMICS OF TENANCY RENT CONTROL 951
In other words, in a competitive equilibrium the nominal rent in period 1 is
(1 ÿ ä)C ˆ R  . In the next month it will be (1 ÿ ä)C=â. The following month
(1 ÿ ä)C=â2 and so on. As long as C < D, all tenants will be leasing in
apartments and the outcome is Pareto optimal.
If we have a rent-control law which does not allow rent-escalation clauses (at
all or adequately) but does allow departure-date contingent rentals, we once
again get optimality. We could then think of the landlord offering contracts
~
like R(1), ~
R(2), . . ., which says that you need make no monthly payment (it is
easy to generalise and allow for some nominally-®xed monthly payment) but if
you leave the apartment after t months you make a lump-sum payment of
~
R(t). In that case, it is easy to see that the competitive outcome is Pareto
optimal.
In reality landlords do often mimic this system. They take initial deposits
from tenants and promise to return part of the money if the tenant leaves
early. The returns are however never quite so ®nely-tuned as in the above
paragraph, for fear of falling foul of rent-control laws.
What was described above as a competitive equilibrium may also be char-
acterised as a Nash equilibrium of a game, as done by Mas-Colell et al. (1995,
pp. 443±50) in the context of Akerlof's (1970) model of adverse selection.
Mas-Colell et al construct a model of Bertrand competition between two
landlords, each of whom can supply an unlimited number of apartments at a cost of C
each. The italicised part of the above sentence is clearly an unrealistic assump-
tion. It is technically necessary because the existence of capacity constraints
can give rise to well-known existence problems.
We get around this problem with a different description of the game. In our
model there are m (potential) landlords where m exceeds the total number of
tenants, t. Each landlord can offer at most one apartment for lease. If she does
so, thenS she incurs a one-time cost of C. Each landlord's strategy set, S, is equal
to fN g [0, 1). If landlord i chooses N , it means she does not offer an
apartment for lease. In that case she does not incur C and her pro®t is zero. If
she chooses a strategy R i 2 [0, 1) it means she offers an apartment for lease.
Her pro®t then depends on R i and other landlords' choice of strategies.
We are essentially looking for a strategy m-tuple (s 1 , . . ., s m )  s which is a
re®nement of a Nash equilibrium in which all the entrants choose the same
strategy. We call this a `uniform Nash equilibrium'. A rental value of R is a
uniform Nash equilibrium if, for some t . 0, m ÿ t landlords choose strategy N
and all the other t landlords choose the same rent R 2 [0, 1) and these
strategies constitute a Nash equilibrium.
The payoff functions of the landlords are assumed to have the following
properties: (a) If among all the landlords who choose to enter, j selects the
smallest rent, then landlord j expects a payoff of V (R j ) ÿ C. (b) If t landlords
enter and all but one of them charge the same rent R and the deviant charges
R9 . R, then the deviant's expected pro®t is ÿC.
(b) is a strong assumption but it mimics the idea the idea of competition. It
is a game theoretic formalisation of the concept of equilibrium intuitively used
in many models of adverse selection, such as Stiglitz and Weiss (1981).
# Royal Economic Society 2000
952 THE ECONOMIC JOURNAL [OCTOBER
Given these assumptions, the uniform Nash equilibria of this m-person game
coincides exactly with the competitive equilibria described above. In Fig. 2 if
the cost is C, the only uniform Nash equilibrium is R  and if the cost is C9 the
only uniform Nash equilibrium is R9.
The only case where we can get multiple equilibria is the non-generic special
case where the cost, C, is such that the horizontal line just touches a peak.
That is, there exists i , n such that V (D=v i ) ˆ C. Barring this special case, a
~ is a competitive equilibrium or a uniform Nash equilibrium if
rental, R,
R~ ˆ minfRjV (R) ˆ Cg:
Though we have in this model assumed a perfectly elastic supply of rental
accommodation, this is not necessary. The method of analysis developed here
can be extended to the case where the supply of rental accommodation is
perfectly inelastic or upward rising with respect to the return on rental. For
reasons of illustration consider the case where the supply of rental accommo-
dation is ®xed at a number S, and C happens to be zero. Let us consider the
case illustrated in Fig. 1. Again, let the total number of (potential) tenants be
t. Hence, there are tpi number of tenants of type i. To illustrate the analysis
with a special case, suppose S ˆ tp2 ‡ tp3 . We know immediately that the only
rents that can possibly qualify as equilibria are the ones in the interval [D=v1 ,
D=v2 ] in Fig. 1 since only in such cases demand equals supply. If however the
rent is too close to D=v1 , in particular less than B, each landlord will have an
interest in under-cutting others for the kind of reason discussed above. Hence,
with a ®xed supply of rental accommodation, all rents in the interval
[B, D=v2 ] comprise equilibria. By considering the case where S 6ˆ ftp3 ,
t( p2 ‡ p3 ), tg we can create equilibria where markets do not clear but this
happens in the special sense of tenants, who are indifferent between leasing an
apartment in the rent controlled area and the outside option, being excluded
from the rent controlled sector.

3. In¯ation and Rents

Our model can be used to derive an interesting testable proposition concern-
ing the relation between in¯ation rates and rents in rent-controlled areas. The
analysis however turns out to more complex than it seems at ®rst sight.
It is intuitively obvious that as the rate of in¯ation rises, (1) the inef®ciencies
of a rent-controlled regime become more acute and (2) the long-stayers
become more unattractive as tenants from the point of view of the landlords
than the short-stayers. (1) and (2) are technical results; we derive these
formally in Lemma 4 and Lemma 5, respectively. It is an implication of these
two lemmas that a higher in¯ation rate will result in a higher real rental rate
for starting tenants. Roughly speaking, the logic behind this proposition is that
as in¯ation picks up, the distortions created by tenancy rent control get
exacerbated; the return earned by landlords drops; and the market responds
by causing rents to rise. This may well result in more tenants being excluded
from the rent-controlled market. In a country like India where the entire
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 953
urban rental market is controlled, this means more tenants will be excluded
from the urban rental market altogether.
Let us postpone further discussion until after the formal results are derived.
Note that in our model, a rise in the in¯ation rate simply means a fall in â,
with zero in¯ation being equivalent to ⠈ 1. Throughout the following analy-
sis we will assume â < 1, that is, we exclude the case of negative in¯ation. To
make explicit the dependence of returns on the rate of in¯ation, we will here
write v i and v(i) as v i ( â) and v(i) ( â), respectively.

Lemma 4
If â9 , â, then v(i) ( â9) , v(i) (â):

Lemma 5
v(i) (â9) v(i) ( â)
If â9 , â, then , :
v i ( â9) v i ( â)
Using the two lemmas it is now possible to derive a testable proposition
concerning the relation between in¯ation and rent. There is no reason to
believe that in¯ation alters the outside options of agents in any systematic way.
So we will take it to be neutral. Hence D and C remain unchanged as the
in¯ation rate changes. Let us also work here with the assumption of `free
entry' of landlords, though that can be altered along the lines of analysis in the
last part of Section 2.
Let us see what happens to the V(R) curve as the rate of in¯ation rises, that
is, â changes to â9, which is less than â. From Lemma 1 we know that the ray
Rv(i) will now be ¯atter for every i.
Next consider each peak of the V (R) line, for instance, the peak at D=v2 in
Fig. 1. Clearly, the height of that peak is given by Dv(2) =v2 . From Lemma 5 we
know that as in¯ation increases (that is, â drops) this peak must get lower. This
is, of course, true for every peak.
From the analysis in the previous section we know that the equilibrium point
on the V (R) curve must be at a point where the horizontal line at height C
®rst hits the V (R) curve from the left. From observations in the above two
paragraphs we know that if the in¯ation rate increases the V (R) curve will
shift right with the new peaks no higher than before. It follows that the
equilibrium monthly rent, R, will rise. Since the equilibrium is the real rent
faced by a starting tenant, what we have just established is the proposition that,
as the in¯ation rate rises, the real rent for new leases goes up in economies in
which there is tenancy rent control.
It is also easy to see that if the rise in in¯ation is suf®cient, the rise in the
real rental can be so sharp as to exclude a whole class of tenants who were
earlier leasing apartments from the rental market. This is not surprising at all
because the inef®ciency of tenancy rent control gets more acute as the
in¯ation rate rises.
# Royal Economic Society 2000
954 THE ECONOMIC JOURNAL [OCTOBER
4. Endogenous Quit Decisions
There are several directions in which one can modify and extend the above
model. We will in this Section consider one involving the endogenisation of
the tenant's type. It is true that some people are inherently prone to moving
and some have transferable jobs. But no matter what the inherent penchants,
people do modify their behaviour depending on the conditions in the rental
market. If in¯ation is very high and a rent-control order holds the nominal
rent constant for sitting tenants, inherently peripatetic individuals may try to
change their ways and stay put in one place, and some people with transferable
jobs may quit such jobs. In this Section we shall try to show that the
endogenisation of tenant types can generate some very interesting results,
including the generic possibility of multiple equilibria. Moreover, the removal
of rent control can result in a uniform lowering of rents.
Let us consider the case where all tenants are innately identical but they can
choose to be one of two types: 1 and 2. The assumption of ex ante identity is
inessential and is made for ease of explanation. Type i changes his apartment
every t i months where t 1 , t 2 . In other words, a tenant chooses to be a short-
stayer or a long-stayer. Again, for reasons of simplicity, let us assume t 2 ˆ 1.
In other words a potential tenant has to decide whether to be a short-stayer or
settle down permanently in a rented apartment. Let us see what happens if we
have the kind of tenancy rent control discussed in the previous Section.
Consider alternative life strategies for the tenant. If a tenant decides on a
career path in which he moves to take up a better job, wherever such
opportunities arise, he will be a short stayer. Let his expected life-time wage-
income in this case be W 1 . If alternatively he chooses a life where he stays in
one place and takes up whatever job he gets in the vicinity, he is a long stayer
and his expected life-time wage-income is W 2 . We assume W 1 . W 2 .
Suppose the market rent is R. A person who decides to be a tenant will
choose to be a short stayer if and only if (W 1 ÿ W 2 ) > (v1 ÿ v2 ). By Lemma 1,
we know that the right-hand term is positive. Hence, there exists a critical rent
size, R, such that if R < R, tenants prefer being of type 1 and if R . R, tenants
prefer to be of type 2. Clearly,

W1 ÿ W2
R (7)
v 1 ÿ v2

Let us, as in the end of Section 2, describe the outcome of the rental market
by thinking of this as a game among the m landlords. As before m . t, where t
is the number of potential tenants. For simplicity assume D is very high; so the
potential tenants always choose to be tenants. A tenant's crucial decision now
pertains to what type he will be. If all landlords charge the same rent, R, each
tenant's choice has already been described. If R < R, each tenant chooses to
be of type 1. Otherwise he is of type 2.
Let us denote this decision by the function T : [0, 1) ! f1, 2g. T (R) ˆ 1 if
and only if R < R. Thus T (R) tells us what type the tenants will be, if there is
only one rent, R, prevailing in the market. Now, consider the case where, with
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 955
all the landlords charging R, one landlord deviates to R9. What can the deviant
landlord expect? As in Section 2, we assume that if R9 . R, she expects to ®nd
no tenants. If R9 , R, tenants will of course come to her, but the question is
what life-style will the tenant choose: short-staying or long-staying? It seems to
us, and this is what we will assume, that all tenants will be of type T (R) ± even
the tenant who rents at rate R9 from the deviant landlord. More formally, in
the language of games used at the end of Section 2, we assume: (c) If t
landlords enter the rental market and all but one of them charges a rent of R
and one landlord charges a rent R9 , R, then the tenants attracted by the
deviant landlord will be of type T (R).
This assumption seems to be intuitively very reasonable. Suppose you live in
a city with a tenancy rent control law in which every landlord, except one,
charges a very high rent. The one exception is your landlord who charges a
low rent of R9. Suppose if every landlord charged R9 you would have adopted
the short-staying life style (involving moving every time you got a better job).
What will you do when only your landlord charges R9? There seems little
rational motivation for you to adopt the short-staying life-style. In fact, since
you know that higher rents exist everywhere else, you will have an extra reason
to stay put where you are.
This assumption is crucial to our model and it is worth clarifying that it is
based on intuition, which is external to our model. We, however, believe that it
is a tenable assumption, and hope that future work will be able to derive this
postulate (c), from more basic axioms.
Fig. 3 considers the case in which all landlords charge the same rent R and
the thick line shows each landlord's expected life-time rental income, V (R).
Note that if R < R  (W 1 ÿ W 2 )=(v1 ÿ v2 ), all tenants are of type 1 and if
R . R, all tenants are of type 2.
Now suppose (as in Section 2) a landlord's cost of leasing out an apartment
is C (as shown in the ®gure). To ®x our attention on the interesting case we
consider one where the horizontal line at C intersects V (R) more than once
(at rents R L and R H ). Unlike in Section 2, here both R L and R H constitute
competitive equilibria or, equivalently, uniform Nash equilibria. It is obvious
that R L is an equilibrium. So consider the case where the market rent is R H .
Landlords' pro®ts are zero, but it seems, at ®rst sight, that an individual
landlord can do better by charging a lower rent±anything between R L and R.
Suppose one landlord does so and charges R 2 (R L , R]. She will have no
problem getting a tenant of course. However the tenant who moves in will not
behave like a type 1 tenant, because if he gives up this tenancy there is no
reason for him to expect that he will again ®nd an apartment for rent R.
Hence, the deviating landlord's expected pro®t will be Rv2 ÿ C. This is non-
positive for R 2 (R L , R]. So no one bene®ts from deviating from R H , which is
a competitive equilibrium. For a formal game-theoretic argument we have to
merely cite assumption (c) above to explain why it does not pay to undercut.
The argument that explains the possibility of multiple equilibria given
tenancy rent control is based on the assumption that there are limits to the
number of apartments a single landlord can offer (for simplicity assumed to be
# Royal Economic Society 2000
956 THE ECONOMIC JOURNAL [OCTOBER
Rv1

Rv2

V(R)

RL R RH R Monthly
Rent, R
Fig. 3. Equilibria with Endogenous Types

one in this paper). If a landlord could under cut R H by offering R9 2 (R L , R]

and supply a large number of apartments at that rent, she may be able to cause
tenants to alter their life's strategy from being long-stayers to short-stayers. In
other words (c) would then no longer hold, and so R H would cease to be a
uniform Nash equilibrium. But in any metropolitan city it does seem reason-
able to assume that no single landlord can cause such a large infusion of
apartments as to induce tenants to alter their life strategy.
Finally, if there is no rent control, clearly then R H will cease to be an equil-
ibrium, because landlords can write type-contingent contracts. So a landlord
can deviate to a rent such as R and make it contractual on the tenant quitting
after t 1 periods. It is easy to see that the only equilibrium is at R L in Fig. 3.
The upshot is that, under tenancy rent control, both R H and R L are
equilibria, whereas under the free contract system only R L is. Suppose rent
control is in effect and it is the bad equilibrium, i.e. R H , that prevails. In
comparison to this, no-rent-control is not only Pareto ef®cient but it is Pareto
dominant. The removal of rent control will result in all rents going down.

5. Extensions: Turnover Costs, Signals and Identity

In this Section we explore two extensions to our model: one in which the
landlord incurs turnover costs, and the other where landlords are able to tell
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 957
something about the expected tenure of a potential tenant by identifying the
group to which the potential tenant belongs.
Turnover costs are common in the rental housing market and typically
consist of cleaning, repainting and renewing worn out appliances upon
vacancy. Let us consider the case where there exists a ®xed turnover cost,
which the landlord has to pay each time a tenant vacates an apartment. In this
case, every time a tenant leaves, the landlord has to incur a transaction cost of
ö (the qualitative results are unchanged whether the turnover costs occur at
the start of a tenancy or the end). Then in (1) we need to add the term ÿä t i ö
to the right-hand side and that will mean (2) would be:
1 ÿ ( âä) t i äti ö
v^i ˆ ÿ : (8)
(1 ÿ ä t i )(1 ÿ âä) 1 ÿ ä t i
The hat on the v i is to mark out this general case. In other words, v^i is the
present value of income earned by a landlord if she charges a nominal monthly
rent of $1 in real terms for every new tenant, gets only tenants of type i and
incurs a cost of ö every time a tenant quits. We de®ne the cost term of (8) as:
äti ö
Öi  (9)
1 ÿ äti
It is easy to see that Ö i , Ö j , for all i . j. Therefore there is now a tension
between the bene®ts to the landlord from a short-stayer, that rents are not
eroded as much as with a long-stayer, and the costs, that turnover costs are
incurred more frequently. This can lead to even more types of tenants being
kept out of the market. De®ne Ö(i) as expected present turnover costs to the
landlord when all tenants of type i or above make themselves available to the
landlord as tenants. We can then write v^(i) as v(i) ÿ Ö(i) . If the rent is R and
only tenants of type i and higher seek tenancy, the landlord's present value of
income from leasing out one apartment is given by Rv(i) ÿ Ö(i) . This follows
from (6), (9) and (13). It is now easy to see that with this speci®cation it is no
longer generally true that if i , j, v^(i) . v^( j) , and so from the landlord's point
of view the shortest stayers are no longer necessarily the ideal tenants.
The tension between the desirability of oft-reset rents and undesirability of
frequent turnover costs can potentially alter the equilibrium from Section 2.
Shorter stayers are likely to be excluded from the rental market by landlords
who want to avoid frequent turnover costs. They accomplish this by setting
rents high enough so that short-staying tenants exit the market. Therefore
turnover costs can lead to the exclusion of tenant types that would not have
been excluded in the absence of these costs.
In reality tenants are not as faceless as assumed in the model. People differ
by race, age, profession, nationality, and so on. Some of these may give a
potential landlord a hint of how long or short-stayer the tenant is likely to be.
A person of a different nationality with a temporary visa cannot be a long-
stayer. Landlords may trust people of a certain age group or a certain race
more. In that case, at the time of leasing they can ask such tenants how long
they plan to stay and put greater trust in the answer than they would with other
# Royal Economic Society 2000
958 THE ECONOMIC JOURNAL [OCTOBER
kinds of tenants. Landlords may also believe that certain groups of tenants are
more likely to leave when asked to do so. They will then prefer tenants
belonging to such a group.
The examples that we have given above are a mixed basket of inherent
characteristics and signals which a tenant can choose (e.g. his profession). For
reasons of simplicity let us consider the case of inherent characteristics.
Suppose people belong to several different races, A, B, C, . . . The expected
length of stay of a tenant differs depending on which type one is and this is
known to landlords. So though, within each category, there are short-stayers
and long-stayers, on average A may be the shortest stayer; followed by B, C and
so on. In such a market what we will expect to see is a variety of rents. These
will be contingent on the tenant's race.
Thus, our model predicts that people in a profession which is known to
involve transfer and moving, or belonging to some identi®able group that is
known to be footloose, will be found paying a lower rent than those who are
known to sink roots in one place and settle down.

6. Conclusion
Rent control laws have been enacted in many countries around the world, mak-
ing them one of the most popular public policy prescriptions among metropo-
litan governments. Unfortunately, knowledge of the effects of tenancy rent
control (which is one of the most pervasive forms of rent control) is inadequate,
especially in the context of positive in¯ation. This paper constructed a model of
tenancy rent control and showed that this kind of rent control system, with
asymmetric information and exogenously given tenant types (the `type' of a
tenant being identi®ed in terms of how long a tenant expects to stay in the same
apartment), lead to outcomes that are Pareto sub-optimal. Free contracting,
however, allows the agents in this model to overcome the asymmetric informa-
tion problem. The paper then studied a model in which how long a tenant stays
in one place is decided by the tenant on the basis of market signals. This
captures the fact that many agents make lifestyle choices depending on the
conditions of the rental housing market. Endogenising the tenant's type gives
rise to the possibility of multiple equilibria in our model. Removal of rent
control laws can not only increase ef®ciency in the rental market, but can also
lead to a general lowering of rents, making all tenants better off.
A number of empirical implications arise from our model. Since landlords
cannot write departure date contingent contracts or have a rent escalation
clause included in the contract, the landlord must set initial rents higher to
compensate for the erosion of real rents suffered during occupancy. This
should lead to across-the-board higher rents in rent-controlled apartments that
are being offered on the market (vacant apartments) than comparable offer-
ings in non rent controlled cities (as found in Nagy, 1997). One would also
expect to ®nd evidence of a tenure discount in rent controlled cities (as in
Nagy, 1997, and BoÈrsch-Supan, 1986), where tenants who have rented the
same apartment for many years pay considerably less in rent than do tenants
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 959
who have only just recently taken possession of an apartment.10 Some recent
research (see Olsen, 1990) suggests that the construction of new housing need
not be any less frequent in a rent controlled city. However, in our basic model,
as certain types are excluded from the rental housing market, the supply of
rental housing in the rent controlled market is likely to fall. In addition, if
increased in¯ation were severe enough to cause the exclusion of even more
types (as in Section 3), then the supply of rental housing would fall as well.
An important implication of the model is that rent control might decrease
the mobility of the labour force. As sitting tenants are reluctant to move from a
rent-controlled apartment, they are less likely to accept a higher paying job in
another city. Therefore, empirically, we would expect to ®nd that the average
tenure of renters is higher in rent-controlled cities (as in Nagy, 1997, and, to
some extent, in Olsen, 1990), and that rent control reduces tenant mobility (as
in Nagy, 1995, and Ault et al., 1994). In fact, Nagy (1997) presents empirical
®ndings that are directly in line with our model. He studies New York City
apartments under the post-1974 rent stabilisation scheme and ®nds that in
1981, rent stabilised apartments had higher initial rents than non-stabilised
apartments. Six years later, for those tenants who remained in the same
apartment in both sectors, those in the stabilised sector paid lower rents. In
addition, tenants in the stabilised sector had longer tenure durations than
tenants in the non-stabilised sector. Thus, the received empirical evidence
generally supports these hypotheses, drawing a picture remarkably similar to
the one that is implied by our model, but the scarcity of detailed empirical
evidence suggests that there is still work to be done in this area.
Our model also drew attention to some systematic relations between rates of
in¯ation and rental rates for property under tenancy rent control. These are
testable propositions but evidence on this is hard to come by. It is hoped that
our model, by clarifying the theoretical link between in¯ation and rent, will
prompt researchers to collect and verify empirically the claims that arise from
this theory.
From the above set of results, it is easy to get the idea that the optimal policy
solution is to free the rental housing market of all government restrictions. We
caution the reader from extending this logic too far, however. As we discuss
earlier in the paper, free contracting in the rental housing market, in the sense
that we use it, does entail certain important responsibilities on the part of
government. The government provides the framework in which contracts are
enforced, and though in our model the absence of rent control was associated
with a system of free contract, there will in reality be three important kinds of
limits on the range of contracts allowed. First, since every society considers
certain kinds of activities illegal, a contract that speci®es the use of some illegal
activity would naturally not be recognised even of both parties voluntarily
agree to it. A contract which entails the landlord killing a tenant who fails to
pay the rent would belong to this category. Second, a contract which adversely

10
Tenure discounts may also occur in non-rent controlled areas as well, however the empirical
evidence is mixed. See Guasch and Marshall (1987).

# Royal Economic Society 2000

960 THE ECONOMIC JOURNAL [OCTOBER
affects an uninvolved `third party' (that is, someone who is not a signatory to
the contract) would be considered illegal. Finally, what one has to keep in
mind is that, in this age of lawyers, contracts can soon become so complicated
that it becomes virtually impossible for the signatories to understand fully what
exactly they are agreeing to. In such a scenario, the more savvy can `trap' the
others into transactions that they would not have got into if they understood
the full rami®cations of the contract. To prevent this from happening, a
practical response is to set some broad limits to the range of permitted
contracts. To the extent that rent controls are themselves partly the conse-
quence of a well-meaning attempt to restrict the range of contracts, one has to
exercise commonsense when limiting the terrain of possible contracts. What
our model suggests is that the class of possible contracts should include rent
escalation clauses, tenancy termination clauses and in general, contracts
involving terms which are departure date contingent. This freedom of choice
allows tenants and landlords to overcome the asymmetric information problem
and reaches not only Pareto ef®ciency but may result in across-the-board lower
housing rentals than what occurs in the presence of rent control.
Cornell University
and University of Colorado at Denver
Date of receipt of ®rst submission: April 1999
Date of receipt of ®nal typescript: April 2000

Appendix A: Proofs of Lemmas

Proof of Lemma 1 We shall, without loss of generality, assume that t j ˆ t i‡1 . Let v kj be
the present value of rents earned by a landlord whose ®rst k tenants are of type i and
all others of type j. (Hence, v0j ˆ v j )
We ®rst show that v1j . v j . Clearly,
v1j ˆ 1 ‡ âä ‡ ( âä)2 ‡


‡( âä) t i ÿ1 ‡ ä t i v j : (11)
Since t j ˆ t i‡1 , and given (3) and (1), we have
v1j ÿ v j ˆ ä t i v j ÿ ( âä) t i ÿ ä t i ‡1 v j

ˆ ä t i b(1 ÿ ä)v j ÿ â t i c
 
âti
ˆ (1 ÿ ä)ä t i v j ÿ :
1ÿä
It is easy to see
âti
vj . : (12)
1ÿä
The right-hand term is the present value of the stream b â t i , â t i ,


c, while v j is the
present value of the sequence b1, â, â2 ,


, â t i , 1, â, â2 ,


, â t i , 1,


c The latter
sequence dominates the former, term by term. Hence, (12) is true, and, therefore,
v1j . v j .
It is easy to check, v kj . v kÿ1
j , 8k, and that lim k!1 v kj ˆ v i . It follows that v i . v j .
# Royal Economic Society 2000
2000] THE ECONOMICS OF TENANCY RENT CONTROL 961
Proof of Lemma 2 Note that, for all k,
v k ˆ 1 ‡ âä ‡ ( âä)2 ‡


‡( âä) t k ÿ1 ‡ ä t k v k
or
1 ‡ âä ‡ ( âä)2 ‡


‡( âä) t k ÿ1 ˆ (1 ÿ ä t k )v k :
Substituting this in (4), we get
0 1
P
n pk tk
BPn C(1 ÿ ä )v k
kˆi @ A
pj
jˆi
v(i) ˆ 0 1
P
n pk tk
1ÿ BPn Cä
kˆi @ A
pj
jˆi

or
P
n
p k (1 ÿ ä t k )v k
kˆi
v(i) ˆ Pn P
n : (13)
pj ÿ pk ä t k
jˆi kˆi

It is worth noting that if the term v k were not there on the right-hand side of (13) then
the right-hand side would be equal to 1. Hence, v(i) is clearly a weighted average of
v i , v i‡1 , . . ., and v n . It is also evident that if j . i, v(i) is obtained from v( j) by
redistributing some of the weights away from v j , . . ., v n to v i , . . ., v jÿ1 . Since, for all
k , j, v k . v j (by Lemma 1), it follows that v(i) . v( j) .
Proof of Lemma 4 From (1) it is obvious that
(1 ÿ ä t k )v k ( â) ˆ 1 ‡ âä ‡ ( âä)2 ‡


‡( âä) t k ÿ1 : (14)
Writing the expression for v(i) derived in the proof of Lemma 2, with the dependence
on â made explicit, we have:
0 1
Pn pk tk
BPn C(1 ÿ ä )v k ( â)
kˆi @ A
pj
jˆi
v(i) ( â) ˆ 0 1 : (15)
P
n pk tk
1ÿ BPn Cä
kˆi @ A
pj
jˆi

By inspecting (14) it is obvious that â9 , â implies v k ( â9) , v k ( â), for all k. From (15)
it follows that v(i) ( â9) , v(i) ( â).
Proof of Lemma 5 Assume â9 , â. From (15) it is clear that if
v k ( â9) v k ( â)
, , for allk . i, (16)
v i ( â9) v i ( â)
then the proof of the lemma is complete. The remainder of this proof is therefore
devoted to establishing (16).
Let k . i. De®ne m  t k ÿ t i . Clearly m > 1. From (2) it follows that
# Royal Economic Society 2000
962 THE ECONOMIC JOURNAL [ O C T O B E R 2000]
v k ( â) 1 ÿ ä t i . 1 ÿ ( âä) t i ‡ m 1 ÿ ( âä) t i ‡ m
ˆ t ‡ m t
Z. :
v i ( â) 1 ÿ ä i 1 ÿ ( âä) i 1 ÿ ( âä) t i
By differentiating the right-hand term with respect to â it can be checked that as â falls,
the term becomes smaller. Hence, (16) is proven.

References
Akerlof, G. (1970). `The market for `lemons': quality uncertainty and the market mechanism.' Quarterly
Journal of Economics, vol. 84, pp. 488±500.
Arnott, R. (1995). `Time for revisionism on rent control?' Journal of Economic Perspectives, vol. 9,
pp. 99±120.
Ault, R. W., Jackson, J. D. and Saba, R. P. (1994). `The effect of long-term rent control on tenant
mobility.' Journal of Urban Economics, vol. 35, pp. 140±58.
Basu, K. (1989). `Technological stagnation, tenurial laws, and adverse selection.' American Economic
Review, vol. 79, pp. 251±5.
BoÈrsch-Supan, A. (1986). `On the West German tenants' protection legislation.' Journal of Institutional
and Theoretical Economics, vol. 142, pp. 380±404.
Cinque, R. A. (1997). `Non-rent protections under rent control and rent stabilization.' New York Law
Journal. June 5.
Drier, P. (1997). `Rent deregulation in California and Massachusetts: politics, policy, and impacts.'
Paper presented at the `Housing `97' Conference at New York University, May 14.
Epple, D. (1998). `Rent control with reputation: theory and evidence.' Regional Science and Urban
Economics, vol. 28, pp. 679±710.
Fallis, G. (1988). `Rent control: the citizen, the market and the state.' Journal of Real Estate Finance and
Economics, vol. 1, pp. 309±20.
Frankena, M. (1975). `Alternative models of rent control.' Urban Studies, vol. 12, pp. 303±8.
Guasch, J. L. and Marshall, R. C. (1987). `A theoretical and empirical analysis of the length of residency
discount in the rental housing market.' Journal of Urban Economics, vol. 22, pp. 291±311.
Hubert, F. (1995). `Contracting with costly tenants.' Regional Science and Urban Economics, vol. 25, pp.
631±54.
Hubert, F. (1996). `Rental contracts, endogenous turnover and rent volatility.' DiskussionbeitraÈge des
Fachbereichs Wirtschaftswissenschaft der Freien UniversitaÈt Berlin.
Jarett, S. and McKee, M. (1997). Rent Regulation in New York City: A Brie®ng Book. New York: Community
Training and Resource Center.
Linneman, P. (1985). `The effect of rent control on the distribution of income among New York City
renters.' Journal of Urban Economics, vol. 22, pp. 14±34.
Mas-Colell, A., Whinston, M. D. and Green, J. R. (1995). Microeconomic Theory. Oxford: Oxford
University Press.
Nagy, J. (1997). `Do vacancy decontrol provisions undo rent control?' Journal of Urban Economics, vol. 42,
pp. 64±78.
Nagy, J. (1995). `Increased duration and sample attrition in New York City's rent controlled sector.'
Journal of Urban Economics, vol. 38, pp. 127±37.
Olsen, E. O. (1990). `What is known about the effects of rent controls?' Consulting Report for U.S.
Department of Housing and Urban Development, Contract Number NA 90-7272, (September).
Raymon, N. (1983). `Price ceilings in competitive markets with variable quality.' Journal of Public
Economics, vol. 22, pp. 257±64.
Smith, L. B., Rosen, K. T. and Fallis, G. (1988). `Recent developments in economic models of housing
markets.' Journal of Economic Literature, vol. 26, pp. 29±64.
Stiglitz, J. E. and Weiss, A. (1981). `Credit rationing in markets with imperfect information.' American
Economic Review, vol. 71, pp. 393±410.
Sweeney, J. L. (1974). `Quality, commodity hierarchies, and housing markets.' Econometrica, vol. 42,
pp. 147±67.

# Royal Economic Society 2000