Professional Documents
Culture Documents
Housing Market Discrimination
Housing Market Discrimination
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
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.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.
http://www.jstor.org
The question of whether discrimination cal studies leads us to conclude that blacks
in the housing market forces Negro house- may pay between 5 and 10 percent more
holds to pay more than white households than whites in most urban areas for com-
for identical bundles of residential services parable housing. Our own analyses of a
has been studied extensively. Still it re- 1967 sample of nearly 1,200 dwelling units
mains a controversial subject. Those who in St. Louis, Missouri suggests a discrim-
claim that discrimination markups exist ination markup in that city on the order of
in urban housing markets rely principally 7 percent.2
on a series of empirical studies which con- Differentials of this magnitude would
clude that blacks pay more than whites for represent a significant loss in Negro wel-
comparable housing or that housing in the fare. However, we contend that re-
ghetto is more expensive than otherwise searchers, in their concern about estimat-
identical housing located outside the ghetto. ing the magnitude of price discrimination,
(See B. Duncan and P. Hauser, R. Haugen have overlooked a far more serious conse-
and A. James Heins, Kain and Quigley Bailey who concludes, "there is no indication that
(1970b), D. McEntire, Richard Muth, C. Negroes, as such, pay more for housing than do other
Rapkin, Rapkin and W. Grigsby, Ronald people of similar density of occupation" (1966, p. 218.)
A detailed presentation of the argument against the
Ridker and John Henning, and M. Sten- existence of a ghetto markup (price discrimination) is
gel.) Those who argue that price discrim- contained in Muth. Muth's own empirical research on
ination does not exist contend that studies the South Side of Chicago, indicates that "Negroes may
pay housing prices that are from 2 to 5 percent greater"
which purport to find evidence of a dis- than whites pay, p. 239. However, he argues that these
crimination markup fail to standardize measured differences, and presumably those obtained
completely for differences in the bundles of in many other empirical studies, are due to cost dif-
ferences (higher operating costs for housing in Negro
residential services consumed by black and neig,hborhoods) and do not represent a discriminatory
white households (see Martin Bailey (1959, markup.
1966), Richard Muth, and Anthony 2 This study, based on a 1967 sample of 629 renter
observations and 438 observations on single-family de-
Pascal).' Evaluation of the diverse empiri- tached housing of the city of St. Louis, goes further than
any previous study in attempting to "standardize" the
* Harvard University and the National Bureau of bundles of residential services consumed by whites and
Economic Research. An earlier version of this paper was blacks. Contract rent or market value (for owner-
presented at the meetings of the American Economic occupied, single-family homes) was regressed upon a
Association in New York, December 28-30, 1969. We detailed set of qualitative and quantitative attributes
would like to acknowledge the assistance of Laura of the bundles of housing services. In addition to vari-
Stieg, Ana Bell, and William McNaught and the ables used in previous studies, such as the number of
helpful suggestions of H. James Brown, Daniel R. Fred- rooms and floor area, our regressions included as ex-
land, Eric A. Hanushek, Clifford R. Kern, Joseph J. planatory variables an elaborate set of quality evalua-
Persky, T. Nicolas Tideman, and Randall D. Weiss. tions for the dwelling unit, the structure, the adjacent
This paper is based on research funded by the National structures, and the immediate neighborhood, as well as
Bureau of Economic Research. The views are, however, indexes of the quality of the neighborhood public school
those of the authors and should not be interpreted as and of the level of criminal activity. Also included was a
reflecting the views of the NBER or any of its sponsors. variable measuring the racial composition of the census
I The only empirical study known to us which fails tract in 1967. These models are discussed in some detail
to find evidence of higher prices in the ghetto is by in Kain and Quigley (1970b).
263
with heads under forty-five years of age; Of primary importance to this discus-
4) couples without children with heads sion is the coefficient of the race dummy
over forty-five years of age; and 5) typical variable. It indicates that, after accounting
families (individuals or married couples for differences in life cycle, income, educa-
with children). tion, and employment status, Negro
Typical families were further described households have a probability of owner-
in terms of age of head, family size, num- ship .09 less than that of whites. Thirty-
ber of school-age children, and by dummy two percent of Negro households in the
variables for: female head of less than sample owned their homes; if they were
forty-five years of age; and female head of white, 41 percent would be homeowners.
more than forty-five years of age. Income, There are some indications thlat the
years of schooling (of head), and number of barriers to Negro occupancy in white
years at present job (for head), were in- neighborhoods are gradually declining.
cluded as explanatory variables for all Thus, it could be argued that current
households. Race was indicated by a ownership patterns primarily reflect his-
dummy variable (1 = Negro, 0 = white). torical discrimination and provide a mis-
The probability of ownership equation, leading view of current conditions. To test
obtained by the method of generalized least this hypothesis, we estimated probability-
squares is summarized by equation (1) of-purchase (1 = purchase, 0= rent) equa-
(Table 1).3 All the coefficients of equation tions for those sample households which
(1) have the anticipated signs and are rea- changed their residence within the past
sonable in magnitude, and twelve are three years. Equation (2) presents the re-
highly significant statistically using con- sults for the probability-of-purchase anal-
ventional criteria. The results indicate ysis (Table 1). The explanatory variables
that old couples are more likely to be are identical to those included in equa-
homeowners than young couples, and old tion (1). The coefficients of the dummy
singles are more likely to be homeowners variables representing household type and
than their younger counterparts. None are age differ in magnitude for recent movers.
as likely to be homeowners as male-headed Aside from these contrasts, the largest
families. Female-headed families are also differences were obtained for the income
less likely to be homeowners than male- and race variables.4 The coefficient of the
headed families. Income and employment race dummy indicates that a Negro mover
are positively related to homeownership. has a probability of purchase .12 lower
Family size is negatively related to home- than an otherwise identical white. Only 8
ownership, but only after adjusting for the percent of Negro movers purchased homes;
different homeowning propensities of f am- had they been white 20 percent would
ilies with school-age children and with ad- have been home-buyers.
ditional workers. The probability of own- Previous levels of housing discrimina-
ership increases as the head of household tion may affect Negro households in at
gets older, and the introduction of a least one important way that is not re-
squared age term yields no evidence of flected in equation (2). Because of past dis-
any significant nonlinearity. crimination, Negro movers are less likely
3 The generalized least squares regression estimates
are obtained by weighting each observation by 4As with the ownership models, separate Negro and
[1/P(1 -P) ]i where P is the value of the probability white equations were estimated for the probability of
predicted by ordinary least squares. It can be shown purchase. Except for their intercepts they were identical,
that this procedure provides more efficient estimates of and a covariance test indicated no statistically different
a linear probability function. relationship.
Household Types
Single females under 45 years
(1 = yes, 0= no) -0.403 -5.596 -0.324 -3.665 -0.191 -2.075
Single females over 45 years -0.295 -5.278 -0.051 -0.569 -0.183 -2.033
Single males under 45 years -0.277 -2.577 -0.283 -2.864 -0.124 -1.170
Single males over 45 years -0.108 -1.207 -0.057 -0.539 -0.196 -1.799
Married couples under 45 years -0.213 -3.407 -0.290 -3.406 -0.095 -1.106
Married couples over 45 years -0.004 -0.070 -0.124 -1.385 -0.111 -1.159
Families
Age of head of household 0.002 2.108 0.004 2.212 0.011 2.527
(Age)2 of head of household -0.000 -2.621
Number of persons (natural
logarithm) -0.156 -3.769 -0.138 -3.342 -0 .113 -3.226
Number of school-age children -0.013 -0.986 0.032 2.626 0.018 1.518
Family headed by female under
45 years (1 = yes, 0= no) -0.007 -0. 148 -0.145 -3. 116 -0. 188 -3. 712
Family headed by female over
45 years (1 = yes, 0= no) -0. 192 -2.561 -0.241 -3.663 -0.206 -3.850
Prior Tenure
Prior owner (1= yes, 0= no) 0.267 4.323
Prior renter (1= yes, 0= no) 0.037 1.315
New household (1= yes, 0= no) -0. 146 -3.945
Intercept 0.409 5.747 0.126 1. 354 0.122 1.212
Degrees of freedom 1166 447 443
R2 0.826 0.301 0.445
than white movers to have been home- size and composition, age, and other mea-
owners in the past. This is important be- sured characteristics. Still, prior tenure
cause when homeowners change their resi- itself may have an independent influence
dence they are more likely to buy than to on subsequent tenure decisions. Therefore,
rent and, conversely, when renters move probability-of-purchase equations were es-
they are more likely to move from one timated with the addition of dummy vari-
rental property to another. ables for prior owner, prior renter, and
In large part the association between new households. A fourth category "prior
past and present, or present and future tenure unreported," is reflected in the in-
tenure arises because renters and owners tercept. Equation 3 in Table (1) illustrates
tend to differ in terms of income, family these estimates.
Both the prior owner and new house- households stratified by the race and by the
hold variables have large and highly sig- years of education of the head of the house-
nificant coefficients. Previous ownership hold (see R. Ramanathan). Presumably,
raises the probability of purchase by .27. this averaging process reduces the transi-
New households are .15 less likely to buy tory component of income and thereby
than are established households of the provides an improved estimate of per-
same age, income, and family characteris- manent income. The ordinary least squares
tics. estimates of the race coefficient for all
Accounting for the effects of prior tenure three equations are consistently larger than
reduces the coefficient of the race variables. those obtained for the current income
Of course, the influence of housing market models as are the GLS estimates for the
discrimination is reflected in prior tenure. purchase model. The GLS estimate of the
In the sample of recent movers, only 2 per- race coefficient in the ownership equation
cent of Negro households had previously using the estimated permanent income is
been homeowners as compared to 17 per- more than twice as large as the race coef-
cent of white households. Yet, even after ficient obtained using annual income, and
controlling for the differences in prior ten- the GLS coefficients in the purchase
ure, Negro households are .09 less likely models using permanent income are also
to become homeowners than white house- larger than those obtained from the equa-
holds in today's "open housing" market.5 tions including annual income (see equa-
Several studies of the demand for hous- tion (1), Table 2).
ing services have concluded that housing An alternative and more dubious (both
expenditures are more strongly related to statistically and theoretically) test of the
permanent than to annual income. By ex- permanent income hypothesis used an es-
tension it might be anticipated that the timate of housing expenditures as a sur-
probabilities of homeownership and pur- rogate for permanent income.6 The most
chase would depend more on permanent obvious statistical problem arises because
than annual income. If this were true, all the estimate of housing expenditures for
or part of the measured difference in the homeowners must be imputed from hous-
probabilities of ownership and purchase ing value using a gross rent multiplier.
of white and nonwhite households in equa- The use of different variables (monthly
tions (1)-(3) might be attributable to un- rent for renters and market value for
measured white/nonwhite differences in owners) transformed by a constant divisor
permanent incomes. in a regression on owner/renter may pro-
As a test of the permanent income hy- duce a spurious correlation. The coef-
pothesis, we followed the convention sug- ficients of this housing expenditure vari-
gested by several authors and replaced able in the ownership and purchase equa-
the annual income term in equations (1)- tions vary between 20 and 45 times their
(3) by the mean incomes of the sample standard errors; this increases our sus-
picion that the relation is to a significant
I A similar difference in Negro-white probabilities of extent spurious and arises by construc-
home purchase was obtained by Daniel Fredland for tion.7
Philadelphia. Fredland's model differed in a number of
respects from equation (3) in Table 1. It included a some-
what different set of explanatory variables, was for 6 This technique was suggested by the anonymous
married households only, and was estimated by ordinary referee. We report it, in spite of strong statistical and
least squares. Even so, he obtained a coefficient of -.16 theoretical reservations.
for a minority dummy (nonwhite or Puerto Rican) and 7 The housing expenditure models reported in Table 2
a coefficient of .27 for the prior-owner dummy. use housing value . 100 as an estimate of homeowners'
Probability of Purchase
Probability of without Prior with Prior
Ownership Tenure Tenure
Equation (1) Equation (2) Equation (3)
OLS GLS OLS GLS OLS GLS
Current Annual
Income -.150 -.088 -.154 -.124 -.114 -.091
(5.06) (2.64) (3.94) (4.55) (2.96) (3.72)
Housing
Expenditure -.048 -.035 - .077 -.048 -.069 -.029
(1.99) (2.33) (2.68) (2.76) (2.35) (1.63)
Taken together the estimates sum- Differences in the asset or wealth posi-
marized in Table 2 and the alternatives tions of Negro and white households may
mentioned strongly indicate that Negro account for part of the differences in white
households are substantially less likely to and nonwhite ownership and purchase
be homeowners or buyers than white probabilities. Unfortunately, the sample
households of similar characteristics. It used in this research shares the deficiency
does not "prove" that this is the result of of most other surveys in not including in-
discriminatory practices in urban housing formation on household assets and wealth.
markets; and there remain several com- Therefore, no direct test of the asset hy-
peting explanations for these results. These pothesis is possible using these data. How-
alternative hypotheses may be grouped ever, for several reasons, we doubt that
into three broad categories: 1) Differences much of the white-Negro differences in
in the "taste" for homeownership be- ownership and purchase are the result of
tween whites and blacks; 2) Differences in an unmeasured difference in wealth. All
the household asset and wealth positions three equations include income, years-on-
of white and black families; and 3) Racial job, and life cycle variables, which may
discrimination in the housing market as account for much of the white-black dif-
the result either of simple price discrim- ferences in assets. For most households,
ination in the owner and renter markets black and white, equity in owner-occupied
or of a more pervasive restriction on the housing is itself the largest component of
supply of owner-occupied housing avail- net worth.10 Therefore, in the probability
able to blacks. "Supply restrictions" of purchase model (equation (3)) prior ten-
could be supplemented or enforced by ure may account for much of the remaining
simple capital market discrimination or by differences in wealth. Down payment re-
an unwillingness on the part of banks and quirements are a major reason why assets
other mortgage lenders to finance home might be expected to affect the decision to
purchases by blacks outside the ghetto. purchase a home. However, FHA and VA
While it is most difficult to prove that down payment requirements, especially
the much lower probability of homeowner- for small single family homes purchased
ship of black households is not due to dif- more than ten years ago, were small or
ferences in the taste for homeownership, nonexistent.
many of the more commonly believed de- Housing market discrimination is the
terminants of the tastes of housing con- third, and to us, the most plausible hy-
sumers are included as independent vari- pothesis explaining the regression results
ables. Furthermore, stratification by race in Table 1. The exact mechanism is hard
for all of the three equations discloses no to specify. Differential price "markups" in
statistically significant differences. We the owner and rental submarkets do not
thus conclude that the "differences in
tastes" hypothesis is not an important ex-
extensive analysis of the mobility rates of the households
planation for the observed differences in in this sample indicates no important differences in the
market behavior between races.9 frequency of moves between white and Negro house-
holds after accounting for other socioeconomic factors.
-.16. When similar analyses were performed for the 10 For example, recent Survey of Economic Oppor-
probability of home purchase, the sample sizes became tunity tabulation indicates that for lower-middle income
uncomfortably small for some subgroups. ($5,000-$7,000 per annum) families, housing equity
9 Household expectations about moving frequency alone represents 40 percent of the net worth of white
may be the only important excluded taste variable. households and an even larger proportion of the net
(As will be discussed subsequently, mobility also affects worth of black households. See the Appendix for further
the economics of homeownership.) However, a fairly details.
explain these differences in Negro and II. Differences Among Metropolitan Areas
white purchase and homeownership prob- A complete test of the supply restriction
abilities.11 We are forced to conclude that hypothesis cannot be accomplished from
"supply restrictions" on Negro residential an analysis of a single metropolitan area.
choice and on the kinds of housing avail- A more powerful test of the effect of supply
able to black households may be largely restrictions can be obtained by analyzing
responsible for the wide discrepancy be- differences in black homeownership among
tween ownership rates for otherwise iden- cities. Metropolitan areas and their ghettos
tical black and white households. differ in terms of the characterisics of their
Further support for this position is housing stocks, and, therefore, in the ex-
provided by data on the average increase tent to which a limitation on being able to
in the market value of Negro- and white- reside outside the ghetto is an effective
owned single family units in St. Louis. For restriction on the supply of ownership-
this sample, the units owned by white type housing available to blacks. For ex-
central city residents have increased in ample, supply restrictions should be much
value at a compound annual rate of 5.2 less important in Los Angeles, where a
percent per year as contrasted to a 7.2 large portion of the ghetto housing supply
percent annual increase for the central consists of single family units, than in
city properties owned by Negro house- Chicago, where ghetto neighborhoods are
holds. If this is interpreted as a differ- predominantly multi-family. We analyzed
ence in the net appreciation of ghetto and the difference between "expected" and
nonghetto properties, the findings of equa- actual black ownership rates in several
tions (1)-(3) become even more difficult metropolitan areas. Expected black own-
to explain. Rather than a difference in the ership rates were computed by multiplying
net appreciation of black and white owned a matrix of white ownership rates (strati-
properties, however, this finding appears fied into income and family size groups) by
to be still another manifestation of limita- the income and family size distribution of
tions on Negro residential choice. White black households. Table 3 presents this
households wishing to improve their measure in 1960 for all eighteen metro-
housing can buy newer or larger houses in politan areas for which the necessary
better neighborhoods. Negro homeowners census data are published.'2 The difference
are much less able to improve their hous- between the actual black ownership rate
ing in this way; as a result we hypothesize and the expected black ownership rate for
that black homeowners spend more for each SMSA is identical in principle to the
renovation and repair than white house- difference in the probability of ownership
holds of similar characteristics. An an- attributed to race in equation (1) for St.
nual increase in suburban white-owned Louis in 1967. (For St. Louis this more
properties of 4.1 percent provides some primitive technique yields -21.0 in 1960
evidence for these inferences. as compared to an OLS estimate of - 15.0
11 Price markups were estimated for owner and renter
occupied structures using the St. Louis sample for three 12 These 18 SMSA's consisted of all those for which
alternative specifications (see Kain and Quigley the data on black and white ownership rates by income
(1970b)). Of the three specifications, two indicate a and family size classes were published. The expected
smaller percentage markup in the owner market. Even black ownership rate was obtained by applying the
if the markup were smaller for rental than for owner- ownership proportions for white households by income
occupied properties, it would require an extremely large and family size for each SMSA to the income and family
price-elasticity-of-choice to reduce the probability of size distribution of black households (see U.S. Bureau
black ownership by 10 percentage points. of the Census (1960a, Table B3)) and summing.
and a GLS estimate from equation (1) of occupied housing or differences in the tim-
-8.8 in 1967.) ing of urban development. Equation (4)
As a test of the supply restriction hy- presents the regression in difference form
pothesis we then regressed these estimated (expected black ownership rate minus ac-
differences upon 1) the proportion of cen- tual black ownership rate), while equation
tral city dwelling units that are single (5) presents the same equation in ratio
family, a proxy for the proportion of the form (expected black ownership rate-
ghetto housing stock that is single family; actual black ownership rate). The t-ratios
2) the proportion of the SMSA black pop- are in parentheses under the coefficients.
ulation living in the central city, a mea-
sure of the extent of suburbanization of (4) (OB- OB) = 0.24 + 0.820w
the black population; and 3) the actual (2.36) (4.64)
- 0.36SC+ 0.12B,
TABLE 3-ACTUAL AND EXPECTED OWNERSHIP
RATES OF NEGRO HOUSEHOLDS BY (6.49) (2.03)
METROPOLITAN AREA
RI = .76
City Actual Expected
(5) (OB/OB) = 0.89 + 1.520W
Atlanta .31 .52
Boston .21 .43 (1.52) (1.47)
Chicago .18 .47
Cleveland .30 .58 - 1.74SC+ 0.90Bc
Dallas .39 .54 (5.34) (2.52)
Detroit .41 .67
Los Angeles/Long Beach .41 .51 R2 = .74
Newark .24 .50
Philadelphia .45 .66 where,
St. Louis .34 .55
Baltimore .36 .61
O* =Expected black ownership rate in
Birmingham .44 .56 the ith SMSA
Houston .46 .56
Indianapolis .45 .58
Memphis .37 .50 [Eawki*Hbki] / Hbki
New Orleans .28 .40
Pittsburgh .35 .59
San Francisco/Oakland .37 .51 OB2= Actual black ownership rate in the
ith SMSA
the kth income/family size cate- occupied units. As the statistics in Table 3
gory in the ith SMSA show, the difference between the actual
S,= Proportion of central city housing and expected homeownership rate of
that is single family (Number of black households is relatively small for
central city dwelling units that cities like Houston and Los Angeles, where
are single family *.total central the central city and its black ghetto in-
city dwelling units) clude more single-family housing, and is
Bc= Proportion of metropolitan Negro relatively large for cities like Chicago,
households residing in central city where the ghetto is predominantely multi-
(Number of Negro households in family and where blacks are effectively
central city *.number of Negro excluded from the suburbs.
households in SMSA). The extent of black suburbanization
also appears to have a significant, though
The means and standard deviations of
small, influence on the gap between actual
the variables used in equations (4) and (5)
and expected black homeownership. In all
are shown in Table 4. The average ex-
U.S. metropolitan areas, black households
are heavily concentrated in the central
TABLE 4-MEANS AND STANDARD DEVIATIONS OF
VARIABLES USED IN INTER-CITY REGRESSION
cities. The mean proportion of blacks re-
siding in the central city for the sample
Mean SD metropolitan areas is .78 and the standard
deviation is only .14. Equation (4) in-
OB-OB 0.19 0.06
0.36
dicates that a city which is one standard
OB/OB 1.61
OW 0.65 0.07 deviation above the mean in terms of this
OB 0.35 0.08 characteristic (92 percent of metropolitan
OB ?0.54 0.07
0.55 0.22
area blacks live in the central city) would
SI
BC 0.78 0.14 have a gap .034 larger than one which is
one standard deviation below the mean
(64 percent of blacks live in the central
pected homeownership rate for black city).
households is .54 and the mean actual The findings presented in equations (4)
black ownership rate is .35. The actual and (5) provide further support for the
white rate for these eighteen metropolitan view that housing market discrimination
areas in 1960 averages .65. Of the .30 dif- limits Negro homeownership.14 Specifi-
ference between actual white and black cally, these results indicate that a limited
ownership rates in these eighteen metro- supply of housing suitable for homeowner-
politan areas Negro-white differences in ship in the ghetto and restrictions on
family size and income account for .11; the Negro purchase outside the ghetto strongly
residual difference, .19, must be attributed affect the tenure-type of the housing con-
to other factors, including the differences sumed by Negro households as well as its
in supply restrictions among the areas. location.
Both equations strongly support the hy-
pothesis that the differences between ob- III. Homeownership, Housing Costs,
and Capital Accumulation
served and expected black ownership
rates are small: 1) when the ghetto hous- Limitations on homeownership have
ing supply includes a larger proportion of significant effects on Negro housing costs,
single-family units; 2) when blacks have
14At the minimum, it would take a peculiar spatial
more access to the suburban housing distribution of tastes for homeownership or of asset dif-
market, with its preponderance of owner- ferences to explain these findings.
tween $10-15,000 in 1962 (see D. S. Pro- tion in urban housing markets has reduced
jector et al.). Negro opportunities for homeownership,
The dominant position of home equities this limitation is an important explanation
in the asset portfolios of low and middle of the smaller quantity of assets owned by
income households is not difficult to under- Negro households at each income level.
stand. Other forms of investment, such as
the stock market, require far more knowl- APPENDIX
housing costs obtained through ownership. (as contrasted with a before tax return of 4
This example assumes that a family may percent on other assets). However, Aaron
choose to buy its dwelling for $20,000 or to implicitly assumes that the real price of
rent it for $167 per month. (This represents owner- and renter-occupied housing is the
a gross rent of $2,000 per year, based on a same. Shelton, in contrast, contends that
widely used gross rent/value ratio). To pur- there is an equilibrium price difference, ex-
chase the unit, the prospective homeowner cluding tax differences, favoring owner-
invests $4,000 as a down payment on the occupied housing by 1.4 percent of value. If
house and assumes a 6 percent mortgage. Shelton's analysis of the comparative costs
As compared to the $2,000 yearly rental of homeownership and renting is correct in
costs, Shelton estimates that purchase would this respect, the savings to homeownership
mean yearly expenses before taxes of $1,590. based on Aaron's example would amount to
Property tax and interest payments create 28 percent of monthly rent computed as
tax shields that reduce the true costs of these [$342+.014 ($25,000 in housing value)]
two items by an amount which depends on ($2,500 annual rent).
the homeowner's tax bracket. He concludes The substantial divergence in housing
that a conservative estimate of the tax sav- costs noted above is in addition to any dis-
ings created by homeownership would be criminatory pricing which may exist. More-
$200, yielding yearly after tax costs of owner- over, it must still be regarded as a lower
ship of $1,390. This represents a saving of bound estimate of the economic cost of an
$610 or a 15.2 percent return (after taxes) effective limitation on homeownership during
on the $4,000 invested in homeownership, as the postwar period, since it fails to incor-
compared to an assumed stock market return porate the effects of inflation on housing costs
of 9 percent before taxes. Since stock market and does not admit to the special position of
earnings are taxable, the comparable before homeownership in the savings behavior and
tax return on homeownership is 18 percent. capital accumulation of low- and middle-in-
The relative return on a homeownership come households.
investment declines as the mortgage is A spending unit's equity in its home can
amortized. The investment return is larger, be divided into three components: the initial
however, if down payments are smaller or if equity or down payment, the amortization
the opportunity cost of equity capital is of the mortgage (savings), and any apprecia-
lower. Thus, 18 percent is likely to be a low tion or depreciation of the property as a
estimate. result of general or particular price changes
The savings from homeownership can also (capital gains or losses). The last two items
be expressed as a percentage of the costs of form the important link between homeown-
renting. From this viewpoint a limitation ership and capital accumulation.
on homeownership would increase housing Although it is technically correct to view
costs beyond three and a half years by 30 an increase in the value of an owned home as
percent, assuming no price appreciation an increase in the household's wealth and to
($610 savings . $2,000 annual rent). As with consider the opportunity cost of the equity
the rate of return analysis, the savings are capital as part of the spending unit's monthly
larger if a smaller down payment or a lower housing costs, there are indications that
opportunity cost of capital is assumed. many households do not view the matter in
Aaron obtains even larger estimates of the precisely this way. Out-of-pocket costs ap-
tax subsidv to homeowners. He presents an pear to be more important considerations for
example, similar to the one just discussed many low- and middle-income families, and
but with a more valuable house ($25,000) it seems many view the savings in the home
and a larger equity ($10,000), which yields a as a bonus to homeownership. Thus, it is of
$342 tax saving (as contrasted to the $200 more than passing interest to compare the
saving computed by Shelton) and an after current out-of-pocket costs of a St. Louis
tax return on a $10,000 equity of 7.4 percent family who purchased an $8,000 FHA or VA