American Economic Association

A Model of Housing Tenure Choice

Author(s): J. V. Henderson and Y. M. Ioannides
Source: The American Economic Review, Vol. 73, No. 1 (Mar., 1983), pp. 98-113
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/1803929
Accessed: 29-05-2017 17:41 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1803929?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
 A Model of Housing Tenure Choice

By J. V. HENDERSON AND Y. M. IOANNIDES*

What are the determinants of tenure choice people choose to own or rent based on how
in the housing market? In the empirical liter- owning distorts their desired path of asset
ature (see, for example, Harvey Rosen; David holdings. Owning may require purchases of
Laidler) at the top of the list for the United assets at a time when dissaving would be
States are the income tax advantages of own- preferred, such as for young people who are
ing. Even given the excessive tax allowances on a low part of their life cycle path of
for depreciation enjoyed by landlords, back- income, or for old people who want to con-
of-the-envelope calculations of equilibrium sume their equity. Artle and Varaiya are able
effective prices indicate that it is still finan- to attain fairly explicit life cycle solutions,
cially advantageous to own rather than rent but at the expense of two unusually restric-
for all income brackets except for, say, the tive assumptions. They assume everyone at a
lowest-income quartile of families, unless a point in time and at every point over the life
family is planning to move soon (see John cycle must consume exactly the same quan-
Shelton). The qualification about moving en- tity of housing, regardless of wealth, price, or
ters because homeowners face higher trans- temporary income differences. Also, it is as-
actions costs of selling and moving than sumed arbitrarily that the opportunity cost
renters. of owning is less than the cost of renting.
While tax laws may favor owning in the The other approach in the theoretical liter-
United States, in many countries the tax ature is to solve for market equilibrium. The
treatment of renting vs. owning is financially only paper we know of which does this is by
unimportant, neutral, or biased towards rent- Yoram Weiss. While his study is interesting
ing. Therefore, in developing a general theo- because there is a complete market solution
retical framework, it is desirable to first con- dividing the world into renters and owners,
centrate on the noninstitutional economic the solution is derived under again unusually
aspects of the problem, and then introduce restrictive assumptions. In a world of perfect
consideration of country-specific institution- certainty, Weiss assumes that people are in-
al factors such as the tax system. This also dexed by differing efficiencies in producing
helps us understand the impact of tax laws housing services from homes if they owner-
on tenure choice. occupy. However, for some reason that ef-
There are two basic approaches to the ficiency is equal for all individuals when they
problem in the theoretical literature. One is rent. Given this dichotomy, not surprisingly
to look at the factors affecting the choices of those whose owner-efficiency parameter ex-
a single consumer. In looking at these fac- ceeds (is less than) the common rental ef-
tors, some recent writers concentrate on the ficiency parameter own (rent), in the absence
risk avoidance behavior of consumers in par- of institutional factors (taxes).
ticular situations facing uncertain future In this paper, we propose to integrate the
rents, housing prices, and rates of inflation key economic elements underlying the papers
(see, for example, John Bossons; Glenn in the literature and to relax the unusually
Canner; loannides). Other writers concen- restrictive assumptions. In particular, any
trate on life cycle aspects of the problem. differences in opportunity cost between rent-
Roland Artle and Pravin Varaiya examine ing and owning will be derived, housing con-
a situation under perfect certainty where sumption will be perfectly divisible and in-
come and price responsive, and there will be
*Brown University and Boston Universitvyuncertainty
respec- about future prices and returns.
tively. We thank an anonymous referee for comments The cost of integration will be some loss in
which were helpful in writing the final draft. terms of explicitness of life cycle and

98

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOA NNIDES: HOUSING TENURE CHOICE 99

market-equilibrium solutions. What we will the attractiveness of owning relative to rent-

be looking at are individuals operating in a ing for individuals with income streams in an
housing market in equilibrium and seeing appropriately defined region.
how their tenure choices are affected by Section III examines tax impacts and
wealth, life cycle, and other considerations. capital market imperfections.
Given this analysis, we can then identify
some of the critical characteristics of an equi- I. The Fundamental Rental Externality
librium.
Before introducing the life cycle and port- There is a basic externality connected with
folio considerations, we explicitly analyze the the rental of a durable that, given equi-
economic differences between the opportun- librium prices, makes it more attractive to
ity cost of renting and owning. This formally own than to rent. To isolate this factor, we
introduces a fundamental ingredient in- turn to analyzing the nature of durable good
volved in tenure choice decisions for any consumption in a world of perfect certainty.
durable, which as far as we know has been The services from a durable are a function
ignored in the theoretical literature. This is of the capacity or stock, and of the rate
accomplished in Section I, where an external- of utilization (see, for example, Guillermo
ity associated with renting durables is identi- Calvo). Where h is capacity and u the rate
fied and shown to be responsible for the of utilization, total services are
relative attractiveness of owning.
Section II introduces uncertainty and deals (1) h = hJf(u); f'>O, f' < o.
with comparisons between owning and rent-
ing when housing may be held as an invest- Thus at the same rate of utilization, services
ment good in a portfolio of assets. The gen- double if we double capacity, but less than
eral proposition is that a person's housing double if we double the rate of utilization for
consumption demand will not equal his the same capacity.
housing investment demand. If a person's The costs of greater rates of utilization
consumption demand for housing is less than include the resource costs of utilization itself
the quantity of the housing stock desired for such as time costs and the costs of increased
investment purposes, consumption demand maintenance and repairs. Events requiring
can be accommodated by simply occupying maintenance or repairs might be broken into
part of one's investment stock. In the ab- rough categories such as (i) breakdowns (for
sence of capital market imperfections, people example, furnace or electrical system in
in this position will definitely owner-occupy a house), (ii) obvious damages (broken
the housing they consume. We show that the windows), and (iii) "hidden" wear and
likelihood of a person being in this position tear (chipped or dirtied paint on walls or
is related critically to the time pattern of scratched floors). All these costs can be sum-
income receipts and the level of wealth. marized by the strictly convex function T(u)
If, on the other hand, in the portfolio where total utilization costs are
problem, consumption demand for housing
exceeds investment demand, a person may (2) h.T(u); T'> O, T"> O.
rent or owner-occupy. In analyzing this, we
assume a person cannot split his tenure so Nontime costs of utilization are incurred as
that part of his consumption is owner- maintenance required to restore the asset to
occupied and part is rented-a person can- its original condition. Alternatively, they
not be in two places at the same time. Thus could represent depreciation (capital loss) of
if he does owner-occupy in this situation, it the asset, or any combination involving par-
means he must "distort" his investment and tial restoration.
consumption choices to bring investment The externality involved in renting arises
levels into equality with consumption levels. from the maintenance problem. An owner-
This "distortion" may be incurred voluntar- occupier directly incurs hcT(u). But a land-
ily because the rental externality enhances lord can collect from the tenant only part of

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
100 THE AMERICAN ECONOMIC REVIEW MARCH 1983

variations in hCT(u), over and above the held constant. The same specification is also
basic contract rent. A tenant, of course, in- used later when uncertainty is incorporated.
curs his own time costs of utilization and, in (ii) For housing consumed in period 1,
addition, the tenant might be charged for costs of utilization are incurred in the second
items under the category of obvious damages period only.'
through, for example, deductions from a If he owns, the consumer's maximization
damage deposit. However, the marginal costs problem is to choose the values (x*, h*, S*,
of increased breakdowns and wear and tear u*} that maximize
caused by increased rates of utilization can-
not be fully charged to the tenant. It is (4) U(x, f ( u) hc) + V(w),
impossible to explicitly provide in rental
contracts for all possible contingencies, subject to y, = x + Phc + S
let alone to even collect on all contingen-
cies provided for. Thus tenants are assumed
w=y2+S(1+r)+Phc-T(u)hc.
to pay less than owners at all rates of util-
ization.
Tenants pay The function U( ) is the utility the consu
derives from the period 1 consumption bun-
dle and V(Q) is the indirect utility function of
(3) hT.(u); T > 0; Tr"'> 0;,
wealth remaining after period 1. Both U( )
and V(.) are assumed to be increasing and
where (u) < T(u) Vu, T'(u) < T'(u) Vu. strictly quasi concave. The term y, is period
This presents a classic externality problem 1 income; x is period 1 consumption of the
in the rental market, where tenants do not numeraire, which is all other goods; hc is
face the social marginal costs of their utiliza- housing stock purchased and u is its rate of
tion rates. This is not the case for owner- utilization; S is period 1 savings which earns
occupiers and is a critical factor in the the market real rate of interest r; P is the
determination of the opportunity cost of constant market purchase price of a unit of
renting vs. owning and occupying. We note it housing stock; and y2 is the income the con-
is also possible to rephrase this discussion in sumer receives in the beginning of period 2.
terms of a tenant being unable to collect Price P can differ between periods without
from a landlord for improvements he makes affecting the results (in equation (6) below, R
on an apartment. adjusts accordingly). We shall use v* to de-
note the maximum value of the consumer's
A. Renting vs. Owning lifetime utility, when he owns.
If he rents, the budget and wealth con-
To isolate the impact of this externality, straints of the consumer's maximization
we examine a consumer deciding whether to problem change. A renter, therefore, chooses
rent or own under conditions of perfect cer- the values {x, hc, , u} that maximize
tainty. Then we state the condition for market
equilibrium and, based upon this condition, (S) U(x, f (u) hc) + V(w),
do a comparison of owning vs. renting. The
following important assumptions are made: subjectto y, =x+S+Rhc;
(i) The consumer maximizes a mul-
tiperiod utility function. However, in terms of W = y2+ S(1 + r)-T(u)hc,
examining optimal period 1 decisions, we
can bury a third and subsequent periods in
the indirect utility function given remaining
'While it might be more realistic to specify time or
wealth at the beginning of the second period.
breakdown costs as being incurred in period 1, to con-
This assumes that optimal decisions are made
serve on notation and the number of functions we
in future periods, and that for our compara- specify them as occurring only once, with no loss of
tive statics, future prices and incomes are generality.

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOA NNIDES: HOUSING TENURE CHOICE 101

where R is the rental price of housing. We

shall use v to denote the maximum value of
lifetime utility when the consumer rents.
T(u)
In market equilibrium, for there to be
owners of rental housing, interest foregone
on the equity in housing must equal housing
\ T (u)
profits. Thus for each unit of housing, equi-
librium in asset holdings requires that

(6) = R _ +(

where ui indicates that this u is the tenant's

MB
and not his landlord's choice variable (al-
though under perfect certainty the landlord
must know what its value will be). Given the
tenant's ui, the landlord directly recovers T(ii) U U U

from the tenant, but must pay T( u-) for

utilization. Given the market-equilibrium FIGURE I

purchase price of housing, for housing to be

held as an asset by a landlord, rents are set
at a level that covers the financial opportun- (this follows from the separability in (1) and
ity cost plus costs of utilization not directly (2)).
recovered from the tenant, all appropriately The left-hand side of these equations can
discounted. Thus although tenants only di- be interpreted as the marginal benefits of
rectly pay T(ui) for utilization, they indirectly increasing utilization. For an owner, the
pay the balance in the form of higher rents. benefits are the increase in the rate of ser-
Finally in equation (6), we could specify vices from capacity, f'(u*), multiplied by
savings for either renters or owner-occupiers the marginal utility of capacity (where
as containing investment in housing. But (U2/U1)(1 + r) = (rP + T(u*))-f(u*)-1). In
given (6), no results would be altered by equilibrium in a world of perfect certainty,
adding these terms on, since under perfect the renter's ui would have to equal the u- of a
certainty from an investment point of view, landlord's tenant in equation (6). By sub-
investment in financial assets and investment stituting (6) into (8), we see that the equi-
in housing are perfect substitutes. librium schedules of marginal benefits of in-
By maximizing utility for renters and creasing utilization as functions of u are the
owners and combining first-order conditions same for (7) and (8). They are graphed in
with equation (6), we can obtain the equi- Figure 1. The marginal benefit curve is
librium rates of utilization under owning u*, downward sloping because f" < 0 (this can
and renting u1, respectively. These are given be seen by differentiating the left-hand side
by of (7) or (8)). The right-hand side of equa-
tions (7) and (8) are the marginal cost of
increasing utilization where for any u T' < T',
(7) MUf ) (rP + T(u*)) = T'(u*), and T", T" > 0. The two curves are depicted
f (U*)
in Figure 1. The obvious conclusion is that
in equilibrium u > uO. The externality results
in overutilization of capacity under rental
(8) M (R(l + r)+ T(u)) ='(u). tenure.
If there are no other considerations, rent-
Note that the level of capacity chosen does ing is always inferior to owning. The variable
not appear in (7) and (8) so that the equi- u is chosen on the basis of T(u) rather than
librium u* and ui are independent of capacity true costs. However, the tenant must indi-

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
102 THE AMERICA N ECONOMIC REVIEW MA RCH 1983

rectly pay the balance in terms of higher individuals from the spectrum of f(.) func-
contract rents (equation (6)). Formally to tions who will rent in equilibrium.2 For that
prove the desirability of owning, for any to occur, we need some other reason for
equilibrium P, we examine v, defined as a renting in the model, such as portfolio con-
Taylor-series expansion about P*. Thus, siderations or capital market imperfections.

v * = U, x X*) +U2* Jh.f II. An Economic Basis for Renting

- h*f (u*))+ V*'(w - w*)+ G. Durables such as housing may serve dual
purposes for many consumers-as a con-
For strictly quasi-concave functions, G < 0. sumption good and as an investment holding
Substituting in, using (6) where u = ui, and in a portfolio. To see this and its implica-
rearranging yields tions in a framework that has become a
standard tool of such investigations (com-
pare Agnar Sandmo, 1968), we examine the
(9) U* <( ) (rP + T( u*))-d} problem of a consumer simultaneously
choosing his optimal housing consumption
and his optimal portfolio.
x fi,(I+r), I< 0 We present the consumer-maximization
problem in a two-period model. However,
The variables 8 and d are defined by the from the work of Paul Samuelson, Eugene
Taylor-series expansions Fama, and Hayne Leland, we also know this
problem of optimal period 1 bundles can be
f(a) = f(U*)+ f'(u*)(a - U*)+ embedded in an n period framework if we
assume all investment and consumer deci-
and T(ui)=T(u*)+T'(u*)(Ci u*)+d, sions from the end of life back to period 1
are made optimally. The function that serves
where, by earlier assumptions on concavity as the period 2 utility function contains as
and convexity, 8 < 0 and d> 0. Owning arguments price distributions in future peri-
dominates renting. ods, future incomes, and future discount fac-
Note that owning dominates renting for tors. As long as these factors are held con-
everyone, even if people have differing ef- stant, we can derive propositions about the
ficiencies in their f( ) functions. One might allocation of wealth among consumption and
think that people who are less efficient in investment goods today. However, it should
producing housing services would rent to try be noted for reference below that the magni-
to shift the costs of high u's to landlords. tudes of relative and absolute risk-aversion
However, such a market equilibrium is not measures are not insensitive to many of the
sustainable, even if landlords cannot dis- factors which have been suppressed in repre-
criminate among potential tenants on the senting future utility by the indirect utility of
basis of differing f(.) functions. The proof is wealth remaining after period 1.
simple. Consider any potential group of peo- In the analysis to follow, we treat housing
ple from the spectrum of f( ) functions who investment as a risky asset, relative to a safe
might rent. For there to be any landlords, a financial asset. This conventional assump-
version of equation (6) must be met where tion may not be the most reasonable one.
contract rents are sufficiently higher than
expected maintenance costs to at least cover 2The referee suggested this situation could give rise
opportunity costs of investment. This means, to a signalling equilibrium where those with lower utili-
however, that people in that group of poten- zation rates would try to communicate this fact to the
landlord. A signalling equilibrium would only be possi-
tial renters with equal to or lower than ex- ble if lower utilization rates are negatively correlated
pected maintenance costs will find it finan- with signalling costs (see A. Michael Spence). We do not
cially disadvantageous to rent, as in equation see the basis for such an assumption in this case; also
(9). Thus there can never be a group of the topic is beyond the scope of the paper.

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOANNIDES: HOUSING TENURE CHOICE 103

Because bonds and

savings deposits in many rents h units of housing capacity for con-
countries (unlike
Israel or Brazil) are not sumption purposes at a price R, and later
inflation-indexed,
their real return may be pays damage payments of T(u)hc, given his
more variable than
the real return to invest- preferred rate of utilization u.
ments in commodities such as housing (or Maximizing (10) with respect to {h., h1, S,
soybeans). However, it should be noted that, ui}, we get
for landlords, even if real capital gains have
a low variance in a stable market, each land-
(I la) -UIR + U2f(a)-E[V']T(a) = 0;
lord may face a high variance maintenance
expenditure (unpaid by the tenant). Below
(lIlb) -(P-R-L)U1 + E[V'(P(1 + 0)
we will explore the impact of altering the
specification of which asset is safe versus
risky. -L(I + r)-(T(u) )-T(ui)))] = 0;

A. Renter vs. Owner-Occupier: (lIc) -U? + E[V'](I + r) = 0;

Behavioral Models
(IlId) U2f'(Cu)-E [ V'] T'( ) = 0.
In the most general case the consumer
chooses his consumption demand, invest- Let v denote the maximum value of lifetime
ment holding of housing, savings, and the utility, (10).
rate of utilization (hI., hjI S, ui} so as to maxi- For the optimal rate of utilization, we
mize again obtain equation (8) by combining
equations (1 la), (1 ic), and (Ild). There-
(10) U(y,-S-(P-L-R)h, fore, u is independent of income, housing
capacity, and portfolio considerations. Other
-Rh., hJf(u))+ E{V( _2 ? S(1 + r)
straightforward manipulations yield

+ (P(1+0)-L(1I -r)-(T(uf)
(12a) f2f (aU) = R+
-( (u)))h1 - T(u .))).

where only income is time-scripted; h, is the

(12b) rP/(I + r) = R
consumer's investment in housing which he
E [ V'(PO -(T(i u)-( Tu)))]
rents out to others in period 1 at a price of
E[V'](1 + r)
R. In period 2 he incurs noncollectable
maintenance costs of (T(iu)- T(iu)) h, that
are uncertain given that the tenants chose u. Equations (12a) and (12b) have standard
He can sell his asset for an unknown return interpretations as marginal conditions de-
of 0. To avoid having to deal with the costs termining h. and hi. In particular, (12b)
of defaults, we assume expresses the standard asset equilibrium con-
dition for assets with stochastic returns of
P(1 + 0)- L(1 + r) P(O)-(T(U)- T(i)), assuming the tenant's u
is uncertain; the return to housing is ad-
-(T(uf)-T(ff))>0, V6,u. justed above the return to the safe asset by
the individual's risk margin.
The term L reflects the fact that he can From (12a) and (12b) it should be clear
obtain a mortgage loan of L at the fixed that in this general problem, optimal housing
market rate of interest r. However, since consumption demand and optimal invest-
(dis)savings, S, is an unconstrained choice ment demand are quite different. If we were
variable for now, at the moment the presence facing a standard consumption-portfolio
of L is irrelevant. In Section III, however, it problem, our specification of the maximiza-
will play a critical role. Finally, the consumer tion problem would end here. We would turn

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
104 THE AMERICAN ECONOMIC REVIEW MARCH 1983

directly to the interpretation of the results subject to the constraint

(Part B), focusing on the issue of when con-
sumption demand for housing exceeds or
(13 Y) hi - hc >, 0.
falls short of investment demand. But in Constraint (13') is indispensable for the
equations (10)-(12), consumption and in- validity of the expression (13). Secondly in
vestment are divorced and there is no issue (13), for own-consumed investment, the con-
of tenure per se, because everyone rents their sumer gives up Rh c in rental income in period
consumption. However, we made assump- 1 and pays utilization costs of T(u)hc in
tions earlier about the nature of housing period 2. On rented-out investment, he pays
which will ensure that some people will want uncompensated utilization costs of (h1 -
to owner-occupy their consumed housing. h c)(T(ii)- T(u)) in period 2.
If consumption demand is less than invest- If we let , be the Lagrange multiplier
ment demand, for example, it would be effi- corresponding to constraint (13'), the opti-
cient for the consumer to owner-occupy that mal consumption bundle {h*, h*, S*, u*)
part of his investment up to hc and rent out must satisfy the following conditions:
the rest, hI - *C. In doing so, he will avoid
the rental externality analyzed in Section I;
(14a) - RUi + f(u*)U2+ Ej[V'(T(u-)
and thus it can be shown that he will be
better off (Section III). If consumption de-
mand exceeds investment demand, however, -(u)-T(u*))] =jj;
we assume he cannot own only part of his
consumption; that is, his consumption ten- (14b) (P-L-R)U1-E[V'(P(I?+)
ure cannot be "split." But to avoid the rental
externality, if hc is near h1, he could "distort" - L(l + r)- (T(u-)- T()))] =I
his investment and consumption choices,
(14c) -U?+(l+r)E[V']=0;
raise h, to hc, and owner-occupy his entire
investment.3
Because consumption tenure cannot be (14d) f'(u*)U2-T'(u*)E[V'] = O;
split and because of the rental externality,
the maximization problem in (10) in fact (14e) ,u =0, h* - h* >O ;
only applies to people who are actually going
to rent their consumed housing. Renters are
(14f) ,>0, h7*-h* =O.
taken from the group of people for whom
hC >h1. All people for whom hc<hI and Equation (14a) indicates that the owner-
occupier by avoiding the rental externality
some for whom h C> h. will own. We turn
to their maximization problem now. "gains back" T(u-)- T(u) per unit of housing
investment used for own consumption.
1. The Owner-Occupier 's Problem If h* > h* so y = 0, the marginal condition
An owner-occupier's lifetime utility maxi- for determining h, is the same as (12b).
mization problem is to choose the values However, the other marginal conditions are
{h*, h*, S*, u*) that maximize different. For h* and u* we have

(13) U(y1-S-(P-L)h1
(15a) U2f(u*)/UI
+ R (h,- hc) hc:f (u*))
/ E[V'(T(ii)-T(u-))] T(u*)
+ E{V(y2 + S(l + r)- T(u)hc E [ V'](1 + r) 1 +r

- (hl - hc)(T()- L (u))

(15b) f (u.) R(1+r)+T(u*)
+ [P(1 + 0)- L(l + r)] h,)},

3This is tantamount to a one-directional indivisibil-

E [V'"(T(u-)-T(au))] ,
ity. E [ V']

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. ] HENDERSON AND IOA NNIDES: HOUSING TENURE CHOICE 105

The opportunity cost of consuming an addi- and only if dy1 + dy2/(l + r) > 0. The time
tional unit of one's housing investment is the path of income is tilted if dy1 - dy2/(l + r)
value of net rent foregone (the first two * 0. If dy, - dy2/(l + r) > 0, the time path is
terms on the right-hand side of (15a), plus tilted towards period 1. If dy1 - dy2/(l + r)
the maintenance cost associated with owner- < 0, it is tilted towards period 2.
occupancy). Equation (15b) is similar to (8). In addition to these definitions, we make
However, if a is uncertain, the optimal u* one critical assumption. The wealth elasticity
now is no longer independent of income and of demand for housing consumption (in the
housing consumption, since the level of con- unconstrained maximization problem (10)) is
sumption chosen affects the uncertain utiliza- positive; that is, dh,/d(y, + y2/(l + r)) > 0.
tion rebate (T(u)- T(U)) from increased In short, we assume housing consumption is
own-consumption and owner-occupancy util- not an inferior good.
ization costs, thus affecting risk considera- The basic results can be derived by ex-
tions. amining the general consumption-portfolio
If ,u > 0 and h I= h c, then the interpreta- problem in (10). In doing comparative statics
tion of the first-order conditions is modified we assume U in (10) is additively separable,
accordingly. Combining (14a) and (14b), we we differentiate equations (1la, b, c) with
get respect to YI and Y2, and solve for changes in
h1, I h, and S. Note ( lId), or its equivalent
(16a) f(u*)U21U, = (P - L) - E [(V'IUI) (8), is not differentiated because ui is inde-
pendent of yl, Y2, and all preference parame-
X (P(i + 9)-L(l + r)-T(u*))]- ters. By differentiating and solving we get

Substituting in (14c) into (16a) yields

(17a) dh, =-U'llIU22
D (1 +f(r)( i))2
(16b) f(u*)U2/U1 = P(r/(l + r))

- E[V'O]/UI + T(u*)/(l + r).

xE[(P +y - O)V"]((I + r) dy, -dy2

Housing consumption, which is equal to in-

vestment, is adjusted until the marginal (17b) dhc = D (R(I + r)2+ (1 + r)r)
evaluation of an additional unit of capacity
equals its marginal cost. Its marginal costs
are marginal utilization costs, T(u*)/(l + r), X (E[V"]E[(f + Y)2V"']
plus the value of expected opportunity costs
(foregone interest, rP/(l + r), less the value
of expected capital gains PE[V'0]/U1).
- (E[(# + -y)V"] )(Y. + + dY

B. Who Owner-Occupies and Who Rents Also we note

Let us turn to the central question of this

section. What types of people are likely to (17c) dS= {T(1+r)Y1U1(R(1+r )2
rent vs. own their consumed housing; that is,
for whom will h1 < hc vs. h* > h*? Using the
above models, we can specify conditions un- + T(l+r))(E[V"( +y)2 ]E[V"]
der which the impact on tenure choice of
differences in total wealth and differences in - E[V"(3 + y)]2)? UIU22f( )2
the slope of the path, or tilt, of the income
stream can be unambiguously stated.
Total wealth consists of two components: xE[V"(3 + y)(3 + -y -

initial income or wealth, yi, and second-

period income Y2. Total wealth increases if x(dy1 - dy2/(l + r)).

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
106 THE AMERICAN ECONOMIC REVIEW MARCH 1983

The auxiliary variables are defined as mand for housing increases as income is
tilted towards period 1. That is, dh/d(yl(l
P - PL(I + r); + r)- Y2) > 0 iff dA/dw < 0. Correspond-
ingly, the portfolio demand for housing de-
y-POa-(T(u) -T(u)); clines as income is tilted towards the future.
(iii) The consumption demand for hous-
a Y2 - (a)h? + S(1 r), ing, holding wealth constant, is unaffected
by the tilt of income. By assumption dh,
(P -L-R)(1+r). increases as wealth increases. (As an aside,
we note a sufficient but not necessary condi-
The determinant D of the Hessian ob- tion for this assumption to hold is that V
tained in differentiating equations (1 a, b, c) exhibits both decreasing absolute risk aver-
is given by sion (dA/dw < 0) and increasing relative risk
aversion (dF/dw > 0). These are also suffi-
D = UlIU22(f( ))2 4(1? y-() _- t cient conditions for D to be negative given
concavity of U and V.)
+ {U1I(R(l + r)+ (u))2 Given these results, the relative attractive-
ness of owner-occupancy vs. renting for indi-
viduals who have identical preferences but
+ U22(f(0)) 2(1 + r)2}
differ with respect to period 1 and 2 incomes
can be demonstrated. We do so graphically
X (E[ V"]E [(1 + y)2V"] in Figure 2 by examining regions of the
(y1, Y2) plane where people rent vs. owner-
-(E[(13 + ) ] ). occupy. By result (iii), differences in Yi and
Y2 which leave total wealth unchanged leave
For a consumer to be at a maximum, the housing consumption demand h C unchanged.
Hessian must be negative definite and a nec- Thus, the loci of points on the (yl, Y2) plane
essary condition for this is that D < 0. for which hc is constant are straight lines
To interpret the results in equations (17), with slope - (1 + r). Assuming an increas-
we utilize common concepts in the theory of ing wealth demand for housing, hc increases
risk aversion. We use the coefficient of the in the northeast direction as marked by
degree of absolute risk aversion (A) where h', h2, h3..
A = - V"/V'. Kenneth Arrow and subse- By result (i), the loci of points on the
quent writers argue that it is most "reason- (y1, Y2) plane for which the portfolio hold-
able" to assume that A declines as wealth ings of housing are the same are straight
rises, or that there is decreasing absolute risk
lines with slope (1 + r), along which dy, =
aversion. Both cases are considered, although dy2 (1 + r) -'. For wealth constant, a tilt from
we will focus on the case of decreasing A. Y2 to yi increases investment demand, if and
For later reference, we will also define only if dA/dw <0 from result (ii). Thus,
the relative risk aversion coefficient F= demand for hi increases in the southeast
-wV"/V'. direction as marked by the lines h', h2, h3 if
Examining equations (17), we can prove and only if absolute risk aversion is decreas-
the following. (Proofs where not obvious are ing. If dA /dw > 0, hI decreases in the south-
in the Appendix.) east direction.
(i) If wealth rises (dy(1?+ r)+ dy2 > ) The curve AXA illustrates a possible locus of
with no change in the tilt of the path of points for which hI = hc as yI and y2 change.
income (dyI(? + r)- dy2= 0), dh' = 0; or the Under the assumption of decreasing absolute
demand for portfolio holdings of housing risk aversion, AA must have slope less than
does not change. Similarly the demand for (1 + r) and greater than -(1 + r), so both hc
safe assets is unchanged. and h, move together. (If there is increasing
(ii) If and only if V exhibits decreasing absolute risk aversion, AA\z would have a slope
risk aversion (dA/dw < 0), the portfolio de- greater than (1 + r) or less than - (1 + r).)

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOANNIDES: HOUSING TENURE CHOICE 107

h3 h'
I
hl'h2Uh3
c c
The observed patterns of tenure choice are
Y2 ~~~~~~~~h1< h2h3 at odds with this completely general result.
The question is, why? Part of the explanation
may be that in correlating wealth and tenure
choice empirically, life cycle considerations
may not be controlled for, so that many of
the assumed low-wealth people are those with
strongly tilted income streams (see below).
Even so, the presumption is that controlling
for tilt, empirically, owner-occupancy rates
increase with wealth. The explanation in our
model must lie in factors other than un-
constrained portfolio-consumption demands,
such as the rental externality, tax laws, and
capital market imperfections. We explore
each of these in turn below.

0 Y1
2. Income Path and Tenure Choice
FIGURE 2
In Figure 2, it should be apparent that at a
given level of wealth there may be both
renters and owners. Assuming dAl/dw <0,
holding wealth constant, renters are those
The AA1 locus divides the plane into re- with lower y1(l + r)/y2's, or those whose in-
gions in a general consumption portfolio come streams are tilted towards the future.
where consumption demand exceeds or falls This would imply, for example, that among
short of investment demand. In our specific the young, renters would tend to be the more
application to housing, people to the south educated because their income streams tend
of 1< would owner-occupy because invest- to be more tilted to the future. Also, those
ment demand exceeds consumption demand. with more inherited financial wealth relative
For those to the north of /A~, their consump- to human capital wealth will have income
tion demand potentially exceeds their invest- streams tilted to the present, and will tend to
ment demand; and they are potential renters own. We do not view these as unexpected
(see subsection 3 for further analysis). results. Thus our real concern is accounting
for the observed wealth effects on tenure
1. Wealth and Tenure Choice choice.
The people to the north of /v/, for a given
Yi (1 ? r )/Y2, are those who have higher 3. Impact of the Rental Externality
wealth, as we move up any h, curve. Thus, Intuitive Analysis. The demarcations in
for any relative time path of income, higher Figure 2 using AA are based on problem (10)
wealth people will be renters. This result does which in fact is specific only to people who
not depend on the nature of risk aversion. actually rent their consumption. We know
With increasing absolute risk aversion, an from Part A of this section that owner-
untilted increase in wealth increases con- occupiers face a revised maximization prob-
sumption demand relative to portfolio de- lem; and this would apply to those who
mand. In that case, those to the east of the would initially desire hi > hC, because of
relevant wo (not shown) would be renters; the maintenance externality. Also those for
again these are high-wealth people. whom hi < hc but hi -_ hc are probably bet-
Similarly the result does not depend on ter off "distorting" their choices, equating hi
the characterization of housing as being the and hC, owner-occupying, and avoiding the
relatively risky asset. The demand for hous- rental externality. This suggests in our il-
ing consumption also rises relative to the safe lustrative solution that AA is only a hypo-
asset (S) with an untilted increase in wealth. thetical locus and that the actual boundary

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
108 THE A MERICA N ECONOMIC RE VIE W MARCH 1983

between renters and owners v

is another curve,
say A'z', which lies above AA. In essence,
the rental externality effectively shifts the
boundary between renters and owners in
favor of owning. V~~~~~~V

In fact, what we expect is a region to the

north of &'A', where everyone rents, a strip
on the plane (with 'A' being the north
boundary) where there is the constraint h*=
h* in equations (14) and (16), and the region 6m3

south of the strip where h* > h* and people

hold housing investments over and above
their own homes. This strip is drawn as = ..
stretching below AA for reasons cited below o y

in the discussion of Figure 3.

The rental externality thus affects the rela-
tionship between wealth and tenure choice, O Y2
in the sense that it stretches out the region of
FIGURE 3
wealth in which people owner-occupy. How-
ever, unless the externality is so costly as to
eliminate renting, we would still expect re- The jump comes because once hc h
gions where, controlling for tilt, the highest for any h C, h , combination, the consumer
wealth people would still rent their consump- could owner-occupy hc of his h, investment
tion. This does not imply that low-wealth and avoid the rental externality. However,
people would be the direct landlords of once owner-occupancy occurs, the correct
high-wealth people, especially since the hold- maximization problem is given in (13) and
ings of a single low-wealth person might not this switch to the correct maximization prob-
provide the housing for a single high-wealth lem again raises utility to m l m 5. Along m 5 m Il
person. Rather low- and high-wealth people from m5 to, say, M6 h,> h*, at m6 h =h*
(who also have asset holdings of housing) where ,u = 0 in (14), and from M6 to mr1 > 0
would pool their money through stock hold- so that h* = h* by constraint.
ings in housing corporations. These corpora- It is not clear in Figure 3 whether mi6,
tions would be the direct landlords of high- where h* = h*, occurs to the left or right of
wealth tenants. m2, where hc= h,. We draw it to the left
Technical Analysis. To understand the im- based upon the following reasoning. Shifting
pact of the rental externality on A\A in Figure to owning at M 2 for the same hc and h,
2, we turn to Figure 3 where we plot indirect would raise "disposable" wealth because of
utility against period 2 income with period 1 the savings on the externality. Given the
income being held constant. We also assume same tilt to the income path, this wealth
decreasing absolute risk aversion, so that effect would raise hc relative to hi so, at
renters are those with low y, /y2's. Indirect that y2' we would expect h* > h*, although
utility is increasing in total wealth. The solid hc =hi. If h* > h* then h* =h* to the left of
curve v that turns into a dashed one at ml as m 2. For this same reason, the &'A' strip in
Y2 declines, plots the maximal value of utility Figure 2 extends south of AA.
for the renter's problem, equation (10). The Thus along the dashed extension of v,
solid part, as v increases beyond m ,, denotes m1m2 where h. < hc it pays people to dis-
equilibrium utility for renters. The solid curve tort their choices and set h,- =hC solving
mi5 m, plots the equilibrium values of v*, problem (14) for the case ,u > 0. The trade-
the maximal value of utility under owner- offs involved in adopting this distortion can
occupancy. The move from mim2 to miM5 be seen by doing a Taylor-series expansion
can be broken into two parts. The plotting of v about p*.
of v has a discontinuity at mi2, where it A Taylor-series expansion reveals three
jumps to the curve m3M4. At M2, hc = hi;factors in the tradeoff made by consumers as

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOANNIDES: HOUSING TENURE CHOICE 109

to whether to owner-occupy or rent.4 The taxable. The tax treatment of interest pay-
first two are obvious. The first is the owner- ments is the same for rental and owner-
occupancy savings from avoiding the mainte- occupied units. In short, the tax system makes
nance externality already documented in owner-occupying cheaper than renting. As-
equation (9). The second is the owner- sume momentarily that the income tax rate is
occupancy portfolio distortion cost of chang- constant so that individual tax burdens are
ing hI to h, which is the increase in invest- proportional. In this case, the impact of the
ment multiplied by the marginal cost of the tax system on tenure choice is identical to
increased risk incurred over and above the that of the rental externality (after removing
market risk premium given. The third factor the uncertainty facing landlords about the
is less obvious. A tenant's choice of the utili- rate of utilization u). Owner-occupancy per
zation rate, iu, may differ from the utilization unit of housing is cheaper than renting by
rate anticipated by landlords, EF[ u]. If i > some constant proportion. In Figure 2, this
E[ ii], there is an additional benefit to renting simply stretches out the region of wealth in
because a portion of the marginal cost of the which people owner-occupy. As with the
renterss excessive maintenance rate is not rental externality, unless the tax advantage is
borne by him. Thus, for example, those whose so great as to eliminate renting entirely, we
f( ) functions are shaped such as to result in would still find a region, where controlling
unusually high ii's will find it advantageous for the tilt of the income path, high-wealth
to rent. The opposite is the case for those people rent their consumption.
whose functions are shaped such that u < An effectively progressive tax system,
E [ i]. however, could eliminate the region of high-
wealth renters, implying everyone would
III. Other Considerations owner-occupy their consumption in the ab-
sence of transactions costs of moving. The
A. Taxes reason is that, although the discrepancy be-
tween desired consumption and investment
Suppose individuals face income taxes. increases with wealth, so does the per unit
Imputed rents to owner-occupiers are tax subsidy to owner-occupied consumption.
exempt while rental income of landlords is There will be a set of pairs of average tax
rates and marginal tax rates (degree of pro-
gressivity) above which renting would be
4We compare utility from renting with utility from eliminated.
owning when h* = h * h. We do a Taylor-series expan-
sion of v about v* and substitute in for i, x*, wv, and w*
from equations (10) and (13), from (14), and from the
B. Capital Market Imperfections
Taylor-series expansions of f and f * and T and f * from
(9). Details are available from the authors upon request. Capital market imperfections can be in-
Rearranging the result gives troduced into our model by imposing the
constraint in all maximization problems that
v - U, < (Sf*U2*/U*- E[ V'/U ]d)hd S > 0. That is, consumers cannot borrow
against future income for current consump-
- E[ V'l Ul ) (( T( u ) - T ( u ) - ( T(u)-T )) tion. However, they can borrow to purchase
durables which can be offered as collateral.
+ c-o)( g[V'/U, ,PO Thus the mortgage loan term L in the previ-
ous sections now has a meaningful role.
- T( U)-T(iU)]- Prem),
Formally, the constraint S > 0 changes all
where Prem is the market-equilibrium risk premium on
previous first-order conditions (equations
housing. When choosing to own, and hence bring h, and
(1 lc) and (14c)) on savings from - U1 +
h, into equality, - cov( ) is the premium required to (l+r)E[V'I]0 to
fully compensate the individual for incurring the in-
creased risk alone (ignoring the other factors). Prem is
(18) - U1 +(I+ r)E[P[V']0, if S*s< 0;
the smaller premium available in the market given the
equilibrium of demand and supply of housing as an
investment good. -U1+(l+r)E[V']<0, ifS*=0.

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
110 THE AMERICAN ECONOMIC REVIEW MARCH 1983

All our analyses would require adjustment to perfection where a minimum wealth level
account for how the shadow price y of the was required to obtain a loan, or there was a
constraint (S > 0) changes as total wealth minimum size requirement on a mortgage
and income stream tilt change. Under rele- (eliminating mortgages for cheaper homes).
vant mortgage loan terms, the principal im- If there are sufficient fixed costs to making
pact of the constraint is to make it less loans (irrespective of loan size), then a ra-
attractive to own, since owning requires risky tional banker could impose such a restric-
investment, when in fact, either total dissav- tion.
ings is desired to increase current consump- How do people behave who are con-
tion or dissavings in the safe asset is desired strained by the imperfections? We examine
(but prohibited) in order to finance purchases only the most interesting case, which applies
of the risky asset. This can be seen by using a to those who are already constrained in
Taylor-series expansion to compare renting owner-occupancy to have h, = h*. We want
and owning as we did in Section II. to know what are the impact of income and
Relevant mortgage loan terms are that, per of loan term changes on the demand for
unit of housing, P - L - R > 0; or P - L> R housing. For example, suppose Y2 increases
so that an owner-occupier requires a greater for y, constant. The resulting increase in
gross cash outlay to consume a unit of hous- total wealth enhances housing consumption
ing than a renter. Thus renting is attractive demand. However, the tilting of income to
since current consumption of all other goods the future lowers total portfolio demand and
is higher than under owning, as is desired increases the desire to dissave.
when y > 0. Why the constraint that P - L For this problem, assuming that the con-
- R > O? If P - L - R < 0, for any potential straints S = 0 and hi = hc remain binding,
landlord, the net outlays on housing invest- the relevant set of necessary conditions are
ments are negative, indicating that everyone one for housing consumption (which equals
will have an infinite demand for housing investment), that is equation (16a), and equa-
investments unless default is costly. Col- tion (14d) for uO. Differentiating these equa-
lateral has little meaning if negative net out- tions with respect to h*, u*, y1, and Y2, for
lays are allowed. dh* we obtain
We consider two issues with capital market
imperfections. What is their impact on tenure
choice and how do people who are con- (19) dh*= D9 [(P-L)U,I(f "U2
strained by the imperfections behave? The
basic wealth effects on tenure choice are
+ h - f,)2U22-T"E[V']+ h(T')2E[VVII)]
unchanged by capital market imperfections,
in a model in which housing consumption _dy
rises with wealth, and loans cover some max- _ Y2 { E [ V"'( + Y )] (f3 "U2 + h (f, )2 U2 2
imizing fraction of investment. For high-
wealth people, if the tilt of their income is
such that S > 0 is a binding constraint, then
- T"E [ V'I]) + T'E [ V"I (hfTU22)}
raising their investment towards their con-
sumption demand to owner-occupy is even where we have used auxiliary variables de-
more costly than it was before. Capital fined earlier; D is the determinant of the
market imperfections would only increase the Hessian obtained in differentiating (16b) and
size of the region in Figure 2 where high- (14d); and D > 0 for a maximum.
wealth people rent. It would also enlarge the Concavity of U, V, and f and convexity of
region of lower-wealth people with incomes T alone ensure that the expression in the
tilted to the future who rent. But, in summary,square brackets of the dy1 term is positive.
this standard specification of capital market An increase in y, increases h which is ex-
imperfections is not sufficient to alter our pected since the consumption (wealth) and
basic types of findings. housing investment effects (increased port-
We could, of course, eliminate low-wealth folio demand where only housing is in the
owners by introducing a capital market im- portfolio) move together. These same condi-

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 1 HENDERSON AND IOA NNIDES: HOUSING TENURE CHOICE 111

tions also ensure that an increase in Y2 upon the rate of utilization, which is as-
decreases h. Therefore, the increase in Y2 sumed to be chosen by occupants, whether
causes a reduction in total portfolio demand renters or owners. Equilibrium in the holding
dominating any other possible effects. This of assets implies that owning housing stock
reduction can only be met by reducing hous- does not differ from holding any other asset
ing investment since S = 0 already. unless uncertain rates of return are intro-
Finally, we note that the impact of chang- duced. With perfect certainty, tenure choice
ing loan terms, L, can be analyzed in the was shown to depend on an important exter-
same fashion. By differentiating (16b) and nality associated with renting. We showed
(14d) with respect to L, we get the rather that the presence of the rental externality
obvious result that an increase in loan size along with equilibrium in asset holding im-
per unit of housing increases the demand for plies that owning dominates renting. In addi-
housing. Consumption and investment de- tion, we showed that the equilibrium rate of
mands go together in this case. The loan utilization of housing stock for renters ex-
reduces the cost of the constraint of no dis- ceeds that of owners, and both are indepen-
savings, and housing consumption demand is dent of all individual characteristics. They
enhanced since more money for general con- depend only on market prices, technologi-
sumption is available in period 1. While this cal characteristics, and maintenance charge
result is obvious, some applications are schedules.
less so. Renting becomes more attractive if hous-
The analysis of a change in L dem- ing is subject to random capital gains or
onstrates the impact of expected inflation on losses and consumers may also invest in a
housing demand. Expected inflation under capital market at a fixed rate of return. In
the traditional U.S. mortgage contract where the model with uncertainty, the advantage
payments are a nominal fixed stream tilts which the rental externality confers on
the housing payment stream towards the owner-occupancy has to be weighed against
purchase period. That is, the real value of the the characteristics of consumers' risk avoid-
down payment and initial mortgage pay- ance behavior. We showed that individuals
ments rise relative to more distant payments. who either have less wealth or who receive
That, in effect, is formally equivalent in our relatively more of their wealth in the begin-
model to a reduction in the loan size, raising ning of their lifetimes will be the net lenders
the initial payment and lowering the later of housing, and vice versa for renters. These
one. Thus, people operating under our con- results hold under the set of restrictions on
straints (S = 0, h* = h*) will experience re- preferences which ensure that the consump-
duced demand for housing under the tradi- tion demand for housing increases with total
tional mortgage contract. Robert Schwab wealth. In the course of developing these
devotes much of his 1982 paper to showing results, we showed that within certain in-
just this. come ranges, because of the maintenance
externality, it pays people to "distort" their
IV. Conclusions investment and consumption choices and
owner-occupy rather than rent.
A model designed to illuminate the dual Finally we examined the impact of tax
role of housing as a consumption and invest- laws favorable to owner-occupancy and
ment good was successful in giving a whole of capital market imperfections on tenure
host of interesting predictions about housing choice. Apart from sufficiently large benefits
market behavior and the determinants of to owner-occupancy from either the rental
tenure choice. Some of these predictions agree externality or average levels of taxation,
with stylized facts and others can be tested which eliminate renting entirely, sufficient
empirically. progressivity of the income tax structure was
In our model, housing stock is used to the only assumption which eliminated the
produce housing services and as an invest- region of high wealth renters. Capital market
ment good. The amount of housing services imperfections inhibited owner-occupancy,
produced per unit of housing stock depends but did not alter the general pattern of re-

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
112 THE AMERICAN ECONOMIC REVIEW MARCH 1983

sults. Finally, we showed that, with capital We shall now examine the signs of the
market imperfections when consumption and terms:
investment demand for housing are con-
strained to be equal, housing consumption E(( + y)(/ + y- )VI
increases with first-period income, decreases
with second-period income, and increases E((/3 + y- ) - "}.
with the proportion of housing cost that can
be mortgaged. For the first term, by multiplying by h, and
by adding and subtracting aE((,/ + y -
APPENDIX )VI ),

We shall prove that under the assumptions E((/3+ ?y)(t + y - V h,

dA/dw < 0 and dF/dw >? 0, the term
+ E((13 + -y - V a

E[V"]E{(f3 + Y)2 V")-(E{(f + y)VII))2 - E((/, + y V") a

is positive. In the course of doing this proof, =E{(,83+ ?y - V`)V h[h1(1 + y)+ a])
the signs of all relevant parts of equations
(17) and the determinant D will be explored. + aE{( -( + y))V")
We add and subtract the term E[V"]E{(3 +
-y)(/ + -y - ()V"} to obtain = E{(/, + -y -)"w)

+ aE((S -(/3 + y)) V").

E[V "]E{(/3 + Y)2V"}- (E{(1 + Y)V,,)2
If we can prove both of these expressions are
+ E[V"]E((f + -y)(3 + -y -)" negative, the proof is complete. By using the
definitions of absolute and relative risk-
- E[V"]E((/ + y)(/ + y - )VI aversion coefficients, we can write the above
expressions as equal to
=E[V"]E((1 + y)(f + y - - -y +))
-E{(/3+y- )V'F}
+ E[V"]E{(/3 + y)(3 + -y' -
- aE{(t-(/- + y))V'A}.
- (E((3 + y)V ))
If we define A and F as the values of A and
F when the state of nature 6 is such that
= E{( + y) V")[ E "]
/ + -y - =0 then the above becomes:

- E((f + y) V"}]
E{(F-F)(3+y-O)V')

+ E[V"]E((/ + -y)(3 + y - + aE((A- A)(t -(/ + ?y))V'),

= E[V"JE ((/ + y)(f + y - )VI given from (lI b) and ( llc) E{(/ + -y- )V')
(-) (-) =0; where 3+?y- >[<] 0, we have A-
A > [ < ] 0, if A is decreasing with w, and
- E((1 + y)V") E((3 + -y -)VI - A - A < [ > 0 O, if A is increasing with w.
(-) (?) Clearly E{(,B + -y-)V"}= O, iff A is con-
stant and is positive [negative] iff A is de-
creasing [increasing] with w. Similarly, where
By definition we /3+-y->[<<]
have that 0, F-F<[>] V" 0, if<F 0,is /3

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms
VOL. 73 NO. 7 HENDERSON AND IOANNIDES: HOUSING TENURE CHOICE 113

nondecreasing with w. As a result, the first of Taxation of Income from Capital, Wash-
the above terms is negative, and so is the ington: Brookings Institution, 1969.
second, if a> 0, and F is assumed to be Leland, Hayne E., "Saving and Uncertainty:
nondecreasing and A is assumed to be de- The Precautionary Demand for Saving,"
creasing with w. Quarterly Journal of Economics, August
1968, 82, 465-73.
, "Optimal Growth in a Stochastic
REFERENCES Environment," Review of Economic Stud-
ies, January 1974, 41, 75-86.
Arrow, Kenneth J., Aspects of the Theory of Rosen, Harvey S., "Housing Decisions and the
Risk Bearing, Helsinki: 1965. U.S. Income Tax: An Econometric Analy-
Artle, Roland and Varaiya, Pravin, "Life Cycle sis," Journal of Public Economics, February
Consumption and Ownership," Journal of 1979, 11, 1-24.
Economic Theory, June 1978, 18, 35-58. Samuelson, Paul A., "Lifetime Portfolio Deci-
Bossons, John D., "Housing Demand and sions by Stochastic Dynamic Program-
Household Wealth," in L. S. Bourne and ming," Review of Economics and Statistics,
J. R. Hitchcock, eds., Urban Housing Mar- August 1968, 51, 239-46.
kets, Toronto: University of Toronto Press, Sandmo, Agnar, "Portfolio Choice in a Theory
1978. of Saving," Swedish Journal of Economics,
Calvo, Guillermo A., "Efficient and Optimal June 1968, 70, 106-22.
Utilization of Capital Service," American , "The Effect of Uncertainty of Sav-
Economic Review, March 1975, 65, 181-86. ing Decisions," Review of Economics Stud-
Canner, Glenn, " Redlining and Mortgage ies, July 1970, 37, 353-60.
Lending Patterns," in J. V. Henderson, ed., Schwab, Robert M., "Inflation Expectations
Research in Urban Economics, Vol. 1, and the Demand for Housing," American
Greenwich: JAI Press, 1981. Economic Review, March 1982, 72, 143-53.
Fama, Eugene F., " Multiperiod Consump- Shelton, John P., "The Cost of Renting vs.
tion-Investment Decisions," American Owning a Home," Land Economics, Feb-
Economic Review, March 1970, 60, 163-74. ruary 1968, 44, 59-72.
loannides, Yannis M., "Temporal Risks and Spence, A. Michael, Market Signaling, Cam-
the Tenure Decision in Housing Markets," bridge: Harvard University Press, 1974.
Economic Letters, 1979, 4, 293-97. Weiss, Yoram, "Capital Gains, Discrimina-
Laidler, David, "Income Tax Incentives for tory Taxes, and the Choice Between Rent-
Owner-Occupied Housing," in A. C. ing and Owning a House," Journal of Pub-
Harberger and M. J. Bailey, eds. The lic Economics, August 1978, 10, 45-55.

This content downloaded from 14.141.12.225 on Mon, 29 May 2017 17:41:54 UTC
All use subject to http://about.jstor.org/terms