2002 V30 4: pp. 635–666
REAL ESTATE
ECONOMICS
Does Homeownership Affect
Child Outcomes?
Donald R. Haurin,∗ Toby L. Parcel∗∗ and R. Jean Haurin∗∗∗
We study the impact of homeowning on the cognitive and behavioral outcomes
of children. Using four waves of a comprehensive national panel data set, we
control for many social, demographic and economic variables previously found
to influence child outcomes. The data are a panel, allowing us to control for unobserved household- and child-specific factors. We use a treatment effects
model to address the issue of possible sample selection bias caused by unobserved variables that influence both the parent’s choice of whether to own
or rent and whether to invest in their children. We find that owning a home
compared with renting leads to a 13 to 23% higher quality home environment,
greater cognitive ability and fewer child behavior problems. For children living
in owned homes, math achievement is up to 9% higher, reading achievement
is up to 7% higher, and children’s behavioral problems are 1 to 3% lower.
Although many economic and social factors influence early childhood cognitive and behavioral outcomes, little is known about the independent effect of
homeownership on children. Our study examines the effects of homeownership on a child’s cognitive and behavioral outcomes. To isolate the impact of
homeownership, we control for many economic, demographic and social factors including local community attributes, household and child characteristics.
We find that the children of homeowners achieve higher levels of cognition and
have fewer behavioral problems.
How might homeownership affect children? We suggest two mechanisms, one
being the stronger investment incentive of owners compared with renters, the
other being greater geographic stability. The investment incentive should result
in a homeowner having a better home environment, and we argue that good
home environments positively impact child outcomes. The greater stability of
homeowners suggests that homeowners will develop greater social capital in
their neighborhood. Also, children will be exposed to a more stable school
environment. We again expect a positive impact on child outcomes.
∗
Ohio State University, Columbus, OH 43220 or [email protected].
Ohio State University, Columbus, OH 43220 or [email protected].
∗∗∗
Ohio State University, Columbus, OH 43220 or [email protected].
∗∗
636 Haurin, Parcel and Haurin
The data set that forms the basis for our analysis is the National Longitudinal
Survey of Youth (NLSY), augmented by the NLSY-Child Data. Our sample
consists of more than 1,000 children, ages five to eight in 1988, who also were
surveyed in 1990, 1992 and 1994. The children’s data are matched with extensive social, demographic and economic data on parents. Parental information
was first collected in 1979 and has been updated annually.
We use a random effects econometric model to estimate the impact of homeownership on the quality of the home environment and the impact of home
environment and homeownership on child outcomes. We recognize that there
may be unobserved factors that influence a household’s tendency to both own
a home and invest in its children. If present, the estimation coefficients will be
biased. In our sample, we find evidence that a selection process is present; thus,
to prevent bias we use a treatment effects model.
We find substantial positive effects of homeownership on the home environment
and we find that increased quality of home environment has a statistically
significant and positive effect on increasing child cognition and reducing child
behavior problems. Further, we find support for the hypothesis that the longer a
parent owns a home, the greater is their children’s cognition and the lower are
behavior problems.
Background
U.S. governmental policy has consistently encouraged individuals and households to become homeowners. The most widely recognized means of influencing the choice between owning and renting is the tax code. Many studies have
documented the positive impact of mortgage deductions and the lack of taxation of homeowners’ imputed rental income on the tendency to own a home
(Haurin, Hendershott and Ling 1988 and Hendershott and Shilling 1982). Estimates of the cost of the mortgage deduction in terms of foregone taxes are
large; for example, the Executive Office of the President’s estimate for 2001
is $61 billion (Executive Office of the President 2000, p. 210).1 Other programs that encourage homeownership include subsidized mortgage interest
payments, reduced down payments, government guarantees of home mortgages, the Mortgage Revenue Bond program, the “rollover” exemption and
other favorable tax treatment of capital gains on housing. In the 1980s and
through 1994, the homeownership rate of U.S. households with head under age
1
Some argue that this amount overstates the cost because it does not account for behavioral responses to the tax law (Follain and Melamed 1998).
Does Homeownership Affect Child Outcomes? 637
45 declined, recovering since then.2 Whether there should be public concern
about the social impact of these changes in homeownership rates of young
households with children depends on knowing the full range of impacts of
homeownership.
Justifications for subsidies for homeowning span political, social and economic
perspectives. One argument has been that homeownership increases wealth
accumulation (Engelhardt 1994, Haurin, Hendershott and Wachter 1996 and
McCarthy, Van Zandt and Rohe 2001). Flavin and Yamashita (2000) used data
from the Panel Study of Income Dynamics (PSID) for 1968–1992 and found
that owner-occupied housing had a 6.6% real return, slightly less than stocks
but greater than bonds or T bills. Other arguments for subsidizing ownership
include the assertion that homeowners are more involved than renters in supporting their neighborhoods; examples include participating in organizations
that support public schools and crime prevention. Another argument claims
that homeowners are better citizens and vote at higher rates. In spite of the frequency of such claims, few well-grounded empirical tests have been conducted
(Rohe, McCarthy and Van Zandt 2000 and Dietz and Haurin 2002).
Our test of the impact of homeownership focuses on two aspects of child outcomes, one being cognition, as measured by normed achievement test scores,
and the other being an indicator of behavioral adjustment. The measures of cognitive achievement are positively associated with contemporaneous and subsequent measures of school achievement and educational attainment (Baker et al.
1993). Our index of behavior problems in children is associated with subsequent
adult antisocial behaviors (Caspi, Elder and Bern 1987, Forgatch, Patterson and
Skinner 1988, Kohlberg, LaCrosse and Ricks 1972 and Mechanic 1980).
Our findings are relevant for public policy. While existing homeownership
support programs target renters in general and especially those who can become
first-time homeowners, we suggest further specificity to target renter households
with young children. We also note that our findings suggest that the gap in
the United States in ownership rate comparing minority households with the
national average results in worse child outcomes for minority children. Finally,
our findings suggest an alternative mechanism to improve public school K–12
achievement compared with standard policies that increase funding levels and
school inputs, or compared with programs such as vouchers or charter schools.
2
For heads ages 35 to 44, the homeownership rate fell from 70% in 1982 to 64.5%
in 1994, subsequently rising to 68.2% in the fourth quarter of 2001. The comparable
changes for heads ages 30 to 34 are 57.1%, 50.6% and 55.0%, and for heads ages 25 to
29 they are 38.6%, 34.1% and 40.8%.
638 Haurin, Parcel and Haurin
Literature on Homeownership and Child Outcomes
Green and White (1997) investigate the effect of parental homeownership on the
probability that a 17-year-old remains in school or drops out. They also study
whether ownership affects the probability that a 17-year-old female has given
birth to a child. Using a probit estimation, they find that parental homeownership
reduces the probability of resident 17-year-old children dropping out or giving
birth. A concern about their approach is the omission of parental wealth and
neighborhood attributes as control variables. The result of this omission could
be a biased coefficient of the homeowner variable because of the correlation
of homeownership rates with wealth (Haurin, Hendershott and Wachter 1996)
and with neighborhood attributes (Rossi and Weber 1996). Another concern is
that Green and White use only current tenure status when their estimations are
based on data from the Panel Study of Income Dynamics or High School and
Beyond because no measure of the past history of homeownership is available.
Thus, they could not test whether the impact of homeownership occurred early
in a child’s life or contemporaneously with the observation of late adolescent
outcomes. It seems likely that a youth’s decision of whether to drop out of
high school would be influenced by more than just the current year’s tenure
status of his or her parents. When they use Census PUMS data, Green and
White include a measure of the duration of ownership and find that the earlier
a family owns a home, the greater the probability that the youth remained in
school.
Another issue faced by Green and White and discussed in more detail later in
this study is the possibility of an unobserved factor influencing both the tendency of parents to own a home and their ability to successfully rear children.
Green and White estimate two bivariate probit models where the first probit
equation models a household’s choice of whether to own or rent, and the two
companion probits separately model the child’s outcome for homeowners and
renters. They find no evidence of correlation of errors between the tenure choice
estimation and the outcome estimations; thus, they conclude there are no biasing unobserved factors. In contrast, our tests for selection bias indicate that
unobserved variables affect both tenure choice and child outcomes; thus, we
use a treatment effects model to yield unbiased estimates.
Our study focuses on the cognitive and behavioral outcomes of young children
rather than 17-year-old youths. Thus, there is greater correspondence between
the timing of measuring child outcomes and measuring explanatory factors.
Our data set is a panel, allowing for more efficient estimation. We include a
large number of control variables including the mother’s family background
factors and factors measured during the child’s first years of life. Finally, we
Does Homeownership Affect Child Outcomes? 639
test for the indirect impact of homeownership through improvements in the
home environment as well as for direct impact.3
Aaronson (2000) builds on the Green and White study by including a measure
of geographic stability in the model. He cites the work of Hanushek, Kain and
Rivkin (1999) who find that greater temporal stability of a household increases
a child’s cognitive performance. When retesting the Green and White model,
Aaronson finds that much of the impact of homeownership appears to work
through the greater degree of stability displayed by homeowners. He also agrees
with DiPasquale and Glaeser (1999), who argue that homeowners have a greater
incentive than renters to invest in their home and neighborhood. This investment
should improve child outcomes. In addition to the positive effect of stability,
Aaronson finds that homeownership has a positive impact on the high school
graduation rate of 19-year-old youths through other unspecified routes, but the
size of the impact varies across specifications.
Boehm and Schlottmann (1999) find that, controlling for other influential variables, the children of homeowners are more likely to be better educated and
to be homeowners 10 years after leaving the parental home. They do not test
for the presence of unobserved parental characteristics that might increase both
the tendency to own a home and invest in their children. We find in our study
that a lengthy list of control variables is not sufficient to eliminate the impact of
unobserved parental characteristics on the estimation of child outcomes; rather,
selection bias must be addressed directly.
Newman, Harkness and Yeung (1998) use a geocoded version of the PSID for
1968–1995 to study amount of education achieved by youths age 20. They
focus on the impact of the neighborhood poverty rate, whether the youth
lived in publicly assisted housing and the amount of mobility of the youth’s
household, but they do not include tenure status in the analysis. They find
a relatively small negative effect on educational attainment if a child, age 6
to 10, lived in a neighborhood with a high poverty rate or lived in assisted
housing.
3
A related paper by Currie and Yelowitz (2000) studies the impact of public housing
on whether resident children are held back in school. Their sample contains 86 families
that live in public housing projects. They find that black male children living in projects
are held back in school less than those living in private housing or receiving vouchers,
but they find no significant effects for white or female children. An obvious concern
with their measure of child outcome is that being held back in school also represents a
choice of the school district.
640 Haurin, Parcel and Haurin
Hypotheses and Research Design
Underlying our model of the relationship of homeownership and child outcomes is an economic argument that because the monetary return is greater,
homeowners are more willing to invest in their homes than are renters. Greater
investment in the dwelling and grounds occurs because homeowners capture
the capital gains accruing from this investment while renters do not. Most prior
studies have found that homeowners’ expenditures on home maintenance and
repairs exceed that of renters, even when controlling for the attributes of the
occupant and the property (Galster 1983 and 1987 as well as Mayer 1981). Two
studies measured the depreciation rates of owned and rented properties using
a hedonic price approach. Shilling, Sirmans and Dombrow (1991) find that
owned property depreciates 0.6% less per year than renter occupied properties.
Gatzlaff, Green and Ling (1998) find a smaller difference of 0.15%.
We argue that better home maintenance is one reason why owners’ homes are
better than renters, ceteris paribus. Improved maintenance reduces or eliminates
the negative impacts of defective water systems, unhygienic conditions, leadbased paint and structural hazards. The expected outcome is that the physical
home environment of owners will be better than that of renters. Prior research
consistently demonstrates that home environments are associated with a child’s
cognitive and social outcomes (Parcel and Menaghan 1990, 1994a and 1994b
and Rogers, Parcel and Menaghan 1991).
Even low levels of lead in a home or yard can affect children’s cognitive and
developmental outcomes. The primary sources of lead encountered by children
are lead-based paints and lead in dust that accumulated from previous use
of lead-based gasoline. The Center for Disease Control estimates there are
1.7 million children with significantly elevated levels of lead in their blood. The
Lead-Based Paint Hazard Reduction and Financing Task Force (1995) presents
a detailed statistical analysis of the likelihood of lead-based paints by housing
type. The incidence is greater in older housing (built prior to 1950) in central
cities. Of all units that are owner occupied, 23.7% were built prior to 1950.
In contrast, 31.5% of rental units were built prior to then. Contamination from
lead-based paints is more likely in housing that is deteriorated, that is, where
the home environment is of low quality. Among all pre-1950 housing units
where children under age six were present, 6.7% of rental units were physically
stressed, while only 4.1% of owner-occupied units were. Among all rental units
with children under age six present, 12.0% had peeling paint, while only 8.5%
of owner-occupied units had peeling paint. There is substantial evidence that
environmental lead contaminates rental units to a greater extent than owneroccupied units, and it is known that lead negatively affects children’s cognition
and health.
Does Homeownership Affect Child Outcomes? 641
Living in a multifamily unit structure is most often associated with renting. Prior
research has shown that residence in a high rise is associated with restricted
play opportunities, social isolation and increased levels of psychological stress
in children (Ineichen and Hooper 1974 and Saegert 1982). Elton and Packer
(1986) find that improved housing quality reduces symptoms of anxiety and
depression among adults. Evans et al. (2000) constructed a measure of housing
quality and used it and income to predict psychological distress in two samples
of individuals. In both cases, better housing quality was significant in reducing
the residents’ level of distress. These studies provide a link between homeownership and better home maintenance and better home quality. They suggest that
residents of owned housing will have less stress and be less isolated. Less stress
should translate into an improved level of emotional support for children in
the home environment. The outcome is expected to be fewer behavioral problems and higher levels of cognitive achievement for children in owner-occupied
units.
Another mechanism by which ownership may affect children is through the impact on the parents’ level of self esteem. Balfor and Smith (1996) use Cleveland
data and find that those households who became owners enjoyed an increase
in self esteem. Rossi and Weber (1996) find that homeownership increases
households’ life satisfaction and happiness. Rohe and Stegman (1994a) follow
households over time, and they find a significant increase in the life-satisfaction
measures of the owner-occupiers after a year and a half of ownership. Rohe and
Basolo (1997) find this effect continues through the third year of ownership.
Increased parental self-esteem should result in greater emotional support for
the owners’ children. Greater emotional support should lead to better cognitive
outcomes and fewer behavioral problems.
Our underlying sociological argument is that social capital is an important
factor in determining child outcomes (Coleman 1988, 1990). Social capital
constitutes social resources upon which family members can draw. In general,
social capital is greater in stable human systems; examples including a stable
marriage or geographic location. Homeownership is relevant for building social
capital because homeowners are less geographically mobile than renters (Lee,
Oropesa and Kanan 1994 and Rohe and Stewart 1996), even after controlling for
other sociodemographic factors. One reason for this difference in mobility is the
much greater economic transaction costs of relocation for homeowners (Haurin,
Hendershott and Ling 1988). A result of the greater expected duration of stay
in the dwelling is that homeowners should be willing to invest more in developing positive relationships with neighbors and their community. Prior empirical
studies have shown that a homeowner’s probability of participating in local organizations is greater than is that of a renter, controlling for other causal factors
(Cox 1982, Baum and Kingston 1984, Rohe and Stegman 1994b and Rossi
642 Haurin, Parcel and Haurin
and Weber 1996). Homeowners’ participation rates in local political activities
are greater than that of renters, controlling for age and socioeconomic status
(Rossi and Weber 1996 and DiPasquale and Glaeser 1999). Studies have found
a positive relationship between homeownership and informal social participation (Hunter 1975 and Jeffers and Dobos 1984) and the level of commitment
to the neighborhood (Austin and Baba 1990). This investment in neighborhood
social capital creates neighborhood networks that may promote positive child
outcomes. Although we do not have direct measures of geographic stability
for owners and renters in the sample, our homeownership variable in the child
outcome equation captures this effect.
Basic Model
The basic model is summarized in the following three equations. The general
approach is the household production function model of Becker (1965). For the
ith child:
Hi T = f 1 (Oi T , Xit ),
(1)
where T is the current period,
Ci T = f 2 (Oi T , Hi T , Zit )
(2)
where t = 1, . . . , T, and
Bi T = f 3 (Oi T , Hi T , Zit ).
(3)
In the first equation, the current homeownership status (Oi T ) and a vector of
control variables (Xi T ) produce the current home environment (Hi T ). Menaghan
and Parcel (1991, 1995) find that home environments are produced with three
types of inputs: parental working conditions, family structure and parental background characteristics. Equations (2) and (3) model a child’s current cognitive
outcome (Ci T ) and behavior (Bi T ). We assume these variables are functions of
the current home environment and current and past values of other explanatory
variables (Zi T ). Our featured test is to determine the impact of current homeownership, but we also test for the impact of the duration of homeownership
on child outcomes.
Our estimation procedure exploits the panel nature of the NLSY79 data. It
is likely that explanatory variables are omitted from each equation; thus, we
include a stochastic error composed of two components: a household- or childspecific error and a purely random error. Rewriting (1)–(3) in a linear form and
including the stochastic components yields:
Hi T = β H HOi T + δ H Xi T + α H i + ε H i T
(4)
Does Homeownership Affect Child Outcomes? 643
Ci T = βc HOi T + κc Hi T + δc Zi T + αCi + εCi T
(5)
Bi T = β B HOi T + κ B Hi T + δ B Zi T + α Bi + ε Bi T .
(6)
In (4), the household-specific error, α H i , reflects unobserved parental abilities and attitudes toward creating the home environment. The mean zero random component is ε H i T . In the cognitive and behavioral outcome equations,
the child-specific errors, αCi and a Bi , reflect unobserved innate abilities or
household-specific influential factors. The components εCi T and ε Bi T are both
mean zero random variables. We assume that the αi are fixed over time and
are mean zero random variables in the sample. Use of this “random effects”
approach allows us to conserve degrees of freedom and is appropriate when the
sample is small relative to the underlying population (Greene 1993, p. 469).
The stochastic properties of our assumed model for the jth equation ( j = H, C,
B) are:
E[α ji ] = 0, E[ε ji T ] = 0, E α 2ji = σa2 , E[ε ji T ] = σε2 ,
E[α ji ε jkT ] = 0
E[ε ji T ε jk S ] = 0
for all i, T and k,
if T = S or i = k, E[α ji α jk ] = 0
if i = k.
We assume that the errors between equations are uncorrelated. To establish
whether this richer formulation of the error structure is important in identifying
the variables that influence child outcomes, we compare the above results with
ones derived from ordinary least squares.4
The vector of Z variables in (5) and (6) includes many other factors affecting
child outcomes. Included in Z are parental age, maternal race, education and
mental ability, family size, and marital stability, as well as child characteristics
such as gender, birth weight and health problems (Parcel and Menaghan 1994b).
We include these variables and neighborhood characteristics as controls.
Unobserved Factors
A significant problem encountered when estimating the effect of homeownership is the possibility of unobserved variables impacting both the tendency
of a household to become a homeowner and its desire to invest in positive
child outcomes. Green and White (1997) discuss a plausible example where
parents can be sorted into two groups: ones with low intertemporal discount
rates who tend to accumulate assets and invest in their children and ones with
4
Greene (1995, pp. 288–291) describes the random effects estimation procedure.
644 Haurin, Parcel and Haurin
high discount rates who tend to neither save nor invest in their children. Because of mortgage-lender-imposed down payment constraints, households who
save have a much higher probability of becoming a homeowner. If this type
of sorting occurs and if OLS is used to estimate Equations (4) through (6),
then the coefficient of the homeownership variable will be biased. Bias may be
encountered whatever the underlying reason for the unobserved correlation of
investing in owner-occupied housing and in children.
The econometric problem is that the sample of homeowning parents may be
selective rather than be a random sample of the population of households. A
solution was described by Barnow, Cain and Goldberger (1981) who name
the problem a “treatment effects” model.5 In our case, the treatment is an
indicator variable for homeowning. Using a probit model, we estimate the
parents’ choice of whether to own or rent (tenure choice). We then estimate
child outcome equations separately for renters and homeowners and test for the
presence of sample-selection bias. The coefficient of the inverse Mill’s ratio
is the product of the standard deviation of the random error in the outcomes
equations and the correlation coefficient of the errors in the outcomes and tenure
choice equations. If statistically significant, there is evidence of selection bias
and a correction procedure should be employed. Unlike Green and White, who
found no selection bias in their sample, we find substantial selection bias; thus,
we use a two-step treatment effects model (Greene 1995, pp. 642–643).
In the first step, we estimate the tenure choice equation where O = 1 if the
parents choose to own:
HOi T = γ H Yi T + αHOi + εHOi T .
(7)
In (7), Y is a vector of explanatory variables related to the ownership decision.
The identification question involves the ability of the Y variables to explain
homeownership but be uncorrelated with child outcomes. While there are some
variables common to both Y and Z, such as household wealth, income and
demographic characteristics, one of the critical variables, the relative price of
homeowning, explains only homeownership. This variable has been shown to be
highly significant and quantitatively important to understanding tenure choice
(Hendershott and Shilling 1982). House price variability depends on the spatial
5
Their application was the analysis of the economic returns to higher education. Barnow,
Cain and Goldberger (1981) noted that youths sort into two groups, those who attend
college and those who do not. However, those who select to attend college may differ
in unobserved ways fom those who choose not to attend. OLS estimation of an earnings
equation containing an indicator variable for college attendance will yield a biased
estimate of the economic returns to attending college if the sample of youth attending
college is not random.
Does Homeownership Affect Child Outcomes? 645
and intertemporal variation in the relative price of a constant-quality unit of
owned housing, mortgage interest rates, expected house price inflation and the
tax rate faced by a household. House prices vary substantially in our sample
because it is national and because there is significant cross-sectional variation
in U.S. constant-quality house prices (Haurin, Hendershott and Kim 1994).
In the second step of the correction procedure, we use (7) to create an instrument
for the homeownership variable that appears in (5) and (6). We expect our
instrument to be strong because of the independent effect of house price on the
tendency to become a homeowner; thus, we avoid the problems associated with
weak instrumental variables (Staiger and Stock 1997).6
Although a number of studies have addressed the econometric issues posed
in treatment effects models, very few have considered those that arise when
the treatment is a time-duration variable. In our case, the duration variable is
the length of time a household has been a homeowner. We extend the treatment model to create an instrument for duration of homeowning using a Tobit
approach, and we use this variable as an alternative to the homeownership
indicator.
Data Set
We use the National Longitudinal Survey of Youth (NLSY79) to provide the
data for parents and the NLSY Child data (NLSY-C) to provide data about
children. The NLSY-C is a survey of the children of NLSY79 mothers (Center
for Human Resource Research 1994). Data about the mothers was collected
beginning in 1979 and was collected annually through the end of our sample
period, 1994. Children in the sample are ages five to eight in 1988.7 Our child
data are from 1988, 1990, 1992 and 1994, a series of years when the cognitive
tests results are directly comparable. The retention rates in both samples are
6
Barnow, Cain and Goldberger (1981) describe two techniques to address the selection
bias created in a problem such as the one we face. Their first technique (probit on
homeownership and separate OLS estimations for owners and renters) was used by
Gren and White (1997) and has the advantage of yielding explicit measures of the
presence of selection bias. We also use this method to identify whether selection bias
is present. The disadvantage is that the impact of ownership on child outcomes must
be derived based on differences between the two OLS equations. The second technique
(instrumental variables in the child outcome equation) has the advantage of directly
producing a measure of the impact of ownership on child outcomes. However, it does
not yield a direct measure of whether there is selection bias.
7
Slightly over 10% of our observations are of multiple children in the same household.
When we estimate the model using this slightly smaller sample, the results are nearly
identical to those reported.
646 Haurin, Parcel and Haurin
excellent (90% of NLSY79 respondents). Our locational data are at the level of
counties.
Dependent Variables
Home environment. The NLSY-C data reports two indexes of the home environment, both derived from the Home Observation for Measurement of the Environment scale (HOME Bradley and Caldwell 1981, 1984a and 1984b, Bradley
et al. 1988, and Elardo and Bradley 1981). The physical environment/cognitive
stimulation HOME scale (HOME-C) is based on the responses to 12 to 15 items,
depending on the child’s age. HOME-C items measure the quality of the building and living space and the amount of materials and time spent on activities
related to cognitive stimulation provided by the family for the child. The emotional support HOME scale (HOME-E) is based on the responses to 12 to 14
items, depending on the child’s age. Items measure the nature of family members’ interactions with the child. Both HOME scales are normed so the average
is 100 with a standard deviation of 15. A percentile score is then derived based
on the assumption that the scores are normally distributed. This scoring method
ensures intertemporal comparability of the HOME scales for the four surveys.
A higher value on either scale implies the child lives in a more supportive home
environment.8
Child cognition. We use two measures of child cognition: reading recognition and mathematical achievement. Both measures are normed scores of
the Peabody Individual Achievement Test (PIAT). The reading recognition
(PIAT-Reading) instrument measures word recognition and pronunciation ability (Baker et al. 1993). These data are converted to percentile scores to insure
intertemporal comparability. PIAT-Math measures mathematics achievement.
The test begins with basic skills such as numeral recognition and progresses to
geometry and trigonometry. The normed scores are converted into percentile
scores. The correlation between PIAT Math and Reading Recognition scores is
about 0.5 (Baker et al. 1993).
Child behavior problems. The NLSY-C includes a measure of children’s
behavior problems based on mothers’ reports of 28 items. Achenbach,
McConaughy and Howell (1987) show that parents’ reports are consistent with
the reports of others, including teachers and mental health practitioners. Assessment items include relatively common behavior syndromes in children;
The Cronbach’s α for HOME-C is 0.67 and for HOME-E is 0.61. This statistic is a
function of the number of items in the scale and the average intercorrelation among the
items (Cronbach 1951).
8
Does Homeownership Affect Child Outcomes? 647
examples include “acting-out,” having a strong temper, demanding attention
depressed-withdrawn behavior, and anxious-distractible behavior. These items
are derived from the 1982 Child Health Supplement to the National Health
Interview Survey (Zill 1988) and were primarily drawn from the Child Behavior Checklist (CBCQ, developed by Achenbach and Edelbrock (1981, 1983).
They have been used since the mid-1960s for measuring and assessing child
behavior problems. Normed scores are created based on data from the 1981
National Health Interview Survey, these data having a mean of 100 and a standard deviation of 15. These normed scores are then converted into percentiles,
with the mean being 62 for the full NLSY-C sample. A higher value of the
index indicates a higher level of behavior problems. Our relatively high sample
mean may be because the NLSY-C mothers are younger than the average in the
United States (Baker et al. 1993).9
Explanatory Variables: Homeownership
We observe the homeownership status of the child’s parents each year beginning
in 1979. These data allow us to test for the impact of not only the contemporaneous measure of homeownership, but also its duration. All households in
our sample have children (an important factor in explaining the probability of
homeownership); thus, our sample’s homeownership rates are relatively high
compared with the age-adjusted national rates.
Control Variables: Economic
Nominal variables including wages, nonlabor income and wealth are deflated
to a common base year, 1994, using the CPI-all item index for urban wage
earners.
Maternal wage. The NLSY79 reports the typical hourly wage rate for working women. We estimate the potential wage earnings for nonworking mothers
using the standard Heckman (1979) technique. We use a maximum likelihood
procedure to jointly estimate labor force participation and the wage equation
(Greene 1995, p. 642). Explanatory variables in the labor force participation
equation include descriptors of the mother’s personal and educational characteristics and descriptors of household characteristics such as the number of
children. Explanatory variables in the wage equation include a measure of the
mother’s score on a standardized achievement test, her race/ethnicity, her education, nine regional indicators and a dummy variable indicating whether the
locality is an MSA. We use the estimated wage rate for all observations in the
9
The Cronbach’s α for the Behavioral Problems Index is 0.88.
648 Haurin, Parcel and Haurin
child outcome equations based on our belief that the predicted value is the best
predictor of a woman’s long-term wage.
Father’s wage. Fathers’ and male partners’ wages are calculated as the ratio
of total earnings in the preceding calendar year to total weeks worked. For
nonworking fathers, a wage is estimated as described above. If no father is
present, the variable is set equal to zero.
Nonlabor income. The NLSY79 reports calendar year income derived from returns on savings accounts, stock dividends, rents, inheritances, public transfers
and other sources. Gifts, such as from parents or grandparents, also are included
in nonlabor income.
Wealth. Wealth is reported annually in the NLSY79, and it includes financial
assets, value of owned home, other real estate, owned businesses, autos and other
durables. Debts also are reported; thus, our measure is of net worth (deflated).
These data have been compared to age-adjusted wealth data in the Survey of
Consumer Finances and found to be similar (Haurin, Hendershott and Wachter
1996).
Control Variables: Sociodemographic
Family size. The number of siblings affects the time and monetary resources
available for each child. Blake (1989), Parcel and Menaghan (1994b) and
Downey (1995) find that as the number of siblings increases, household resources are diluted, thereby negatively impacting child outcomes.
Maternal marital history. Mother’s marital status and history have been found
to influence child outcomes (Haurin 1992 and Rogers, Parcel and Menaghan
1991). The mother’s marital history is described by four dummy variables.
The omitted category contains mothers who were married throughout the duration of the child’s life. Dummy variables in the estimation include: (1) single
for the duration of the child’s life (SINGLE), (2) single during the birth year,
married subsequently (GET MARRIED), (3) married during the birth year, divorced/separated/widowed once or more subsequently, not remarried at the
survey date (MARITAL BREAKUP) and (4) married during the birth year, divorced/separated/widowed once or more subsequently, remarried at the survey
date (REMARRY).
Maternal background characteristics. We include eight background characteristics of the mother as control variables in the child outcome equations. They are
(1) the ethnicity of the mother including Mexican Hispanic, Other Hispanic, or
black, with the residual category being non-Hispanic, non-African American,
(2) the mother’s age and its square, (3) the mother’s highest grade completed
Does Homeownership Affect Child Outcomes? 649
(HGC), (4) the mother’s mental ability measured by her age-adjusted normed
score on the Armed Forces Qualification Test (taken by all NLSY79 participants in 1980),10 (5) the level of religiosity (Haurin and Mott 1990),11 (6) the
type of household in which the mother resided when she was age 14,12 (7)
maternal mastery defined by a scale constructed from four items taken from
Rotter’s (1966) locus of control measure13 (this measure is drawn from the
1979 interview; thus, maternal mastery is measured at or about the time these
women gave birth to the child we study) and (8) the mother’s number of paid
hours of work during the child’s first three years of life (TOTAL HOURS MOM
WORK-YRS. 1–3).
Paternal background characteristics. Because a child’s father (or mother’s
partner) is not a NLSY79 respondent, we have less background information
about him. His age and highest grade completed (HGC) are available if he
resides in the household.
Child characteristics. Gender of the child has been found to influence a child’s
cognitive score (Parcel and Menaghan 1994b). We include an indicator of
whether the child has a health condition that limits school attendance, play,
or sports activities. Low birth weight is represented by a dummy variable that
indicates whether the child’s birth weight was below four pounds (Mott 1991).
Control Variables: Community Factors
Neighborhood characteristics have been found to influence youths’ outcomes.
Crane (1991) finds evidence of a nonlinear impact of neighborhood quality on
adolescent outcomes but Clark (1992) could not reproduce the results using
different data. Dornbusch, Ritter and Steinbert (1999) use census track data and
10
The AFQT consists of the sum of scores on four subtests of the Armed Services
Vocational Aptitude Battery, including word knowledge, paragraph comprehension, numeric operations and arithmetic reasoning. Details are provided in Baker et al. (1993).
11
We include five dummy variables showing the mother’s frequency of church attendance. The omitted category is no attendance, the dummy variables are Church AttendLow (up to once per month), Church Attend-Some (two to three times per month), Church
Attend-Often (once per week) and Church Attend-High (more than once per week).
12
We use a series of three dummy variables to define cases where the child’s mother
was living with both parents when she was age 14 (omitted case), was living with her
mother and no other man (MOM-ALONE), was living with her mother and some other
man such as a stepfather or other male relative (MOM-PAIR) and was living in some
other arrangement such as with only her father (MOM-OTHER).
13
The Rotter scale assesses the degree to which a woman feels that she has control over
the direction of her life, she can follow through with the plans she makes, she can get
what she wants without relying on luck and she has influence over the things that happen
to her. A higher value on the scale indicates a higher degree of control.
650 Haurin, Parcel and Haurin
they find that a neighborhood’s socioeconomic status affects the self-reported
grades of high school students in California and Wisconsin. Duncan (1994)
also uses census track data to measure neighborhood characteristics and finds
evidence that affluent neighbors affect youths’ educational attainment, but the
effects differ by race and gender. Some of his findings are quite precise, such
as black males benefit only if the affluent neighbors are black.
Evans, Oates and Schwab (1992) note that neighborhoods and peer groups are
selected by households; thus, neighborhood characteristics are not exogenous.14
In a simple model that does not account for this selection bias, there is evidence of peer effects, but these effects disappear once the selection process
is accounted for. Crane (1991), Dornbusch, Ritter and Steinberg (1991), Clark
(1992) Duncan (1994), and most other empirical studies of neighborhood effects
do not account for endogenous sorting by neighborhood. Thus, their findings
regarding neighborhood effects are unreliable. Briggs (1998) argues that existing studies of neighborhood effects do not test theories, they only represent
correlations. He argues that the important variable is the size and type of social
network established by a household. A neighborhood may not affect a resident household if the household is not connected to the neighborhood. Haurin,
Dietz and Weinberg (2002) conduct a general review of the literature about
neighborhood effects.
We use county-level variables to measure the impact of neighborhoods on child
outcomes because county data are the smallest geographic level of observation
publicly available in the NLSY79.
The source of the county data is primarily various City and County Data Books.
Community variables include median household income, population density,
percent black, percent Hispanic, unemployment rate, poverty rate, crime rate
and average level of education (percent high school graduates and percent with
some college). Clearly, our controls for neighborhood effects are very weak
because their spatial size is too large. Also, we do not attempt to address the
endogeneity issue.
Results
Descriptive Statistics
The number of observations is the same each of the four years in the panel data
set, 1,026 children, yielding 4,104 total observations. Variable means are listed
in Table 1. The correlations of homeownership with the two HOME indexes
14
Other studies addressing this point include Case and Katz (1991) and Haveman and
Wolfe (1995).
Does Homeownership Affect Child Outcomes? 651
Table 1 Variable means.
Variable
Mean
Variable
Mean
Homeownership Rate
Duration of Homeowning
HOME: Cognitive/Physical
HOME: Emotional
Black
Mexican Hispanic
Other Hispanic
Health Limit
Low Birth Weight
Mother’s HGC
Mother’s Age
Mother’s AFQT
Church Attend-Low
Church Attend-Some
Church Attend-Often
Church Attend-High
TOTAL HOURS MOM WORK-YRS.
1–3 (000)
Male
Crime Rate Index
% Hispanic
% Poverty
Median Community Income ($000)
% Black
0.40
2.37
47.15
46.87
0.35
0.03
0.02
0.04
0.07
11.77
31.00
31.16
0.25
0.24
0.23
0.11
1.68
PIAT: Mathematics
PIAT: Reading
Behavior Problems
MOM-ALONEa
MOM-PAIRb
MOM-OTHERc
Maternal Mastery
Mother’s Wage
Siblings
Father’s Aged
Father’s HGC d
Father’s Waged
Nonlabor Income ($000)
Wealth ($000)
SINGLE
GET MARRIED
REMARRY
46.27
54.37
65.77
0.20
0.10
0.09
2.27
8.23
1.67
34.83
12.27
12.78
5.30
4.70
0.17
0.10
0.26
0.50
58.61
9.20
10.91
19.12
13.89
MARITAL BREAKUP
Population Density (000)
Unemployment Rate
% High School Educated
% College Educated
0.12
16.28
7.24
49.38
14.66
a
Mother raised by her mother.
Mother raised by her mother and another man.
c
Mother raised by some other combination.
d
The mean is for only fathers or partners present in the household.
b
are positive (0.28 for cognitive and 0.32 for emotional), as are the correlations
with the cognitive test scores (0.21 for math and 0.19 for reading recognition).
The correlation of homeownership with the Behavior Problems Index (BPI)
is negative (−0.07); this sign is expected because a higher value of the index
indicates greater behavioral problems.
Regression Results
We estimated the mother’s and father’s wage equations for 1988–1994 using
maximum likelihood. The sample’s mean estimated wage of the mother (deflated) is $8.23. The mean estimated wage of the father (if one was present in
the household) is $12.78.15
15
Complete estimation results of the wage equations are available from the first author.
652 Haurin, Parcel and Haurin
The next estimate is of (7), the equation that models the choice of a household
to own or rent. Demographic explanatory variables include the parents’ ages,
race/ethnicity, number of children in the household, age of the surveyed child,
population density and whether the respondent lives in an MSA. Economic explanatory variables include parents’ wages, wealth, an index of house prices,16
mortgage interest rates (current and lagged) and an indicator of whether the
household has sufficient wealth to be able to make a standard down payment
on its desired house.17 The results, listed in the Appendix, have good explanatory power with all variables having signs consistent with those found in the
literature on homeownership. Specifically, the index of housing prices and the
indicator of the level of difficulty meeting the down payment constraint are
large, negative and statistically significant. Of the 4,104 observations, housing
tenure is correctly predicted in 79% of the cases. The Appendix also reports
the results for the Tobit estimation of the duration of homeownership.
We next test for the presence of sample-selection bias. From the results of the
tenure choice estimation, the inverse Mill’s ratio is computed for both owners
and renters (Greene 1995, pp. 638–641). We test for selection bias in six equations: separately for homeowners and renters and for each of the three child
outcomes. Statistically significant selection bias (5% level) is found in all three
renter estimations and in one owner estimation. Bias is most significant in the
behavior problems equation.18 Given this strong evidence for the presence of
selection bias, we use the treatment effects model to measure the influence of
homeownership on home environment and child outcomes.
16
Shelter costs for constant-quality owned housing are derived from the Freddie Mac–
Fannie Mae (FF) repeat sales house price index (Office of Federal Housing Enterprise
Oversight 1998), augmented by data from the American Chamber of Commerce Research Association (ACCRA 1993). The FF index covers more than 100 MSAs and
all 50 states and is a pure time-series price index. We use the 1982 ACCRA data for
88 MSAs and all 50 states to develop a baseline cross-sectional price index. The final
index is developed by applying the FF index to the ACCRA data, yielding a nominal
house price index with excellent spatial coverage. A comparable price series for rental
housing is not available; however, variations in the relative price of owned housing are
dominated by spatial variations in the price of owned housing.
17
This type of mortgage constraint variable has been shown to be important in explaining
the tenure choice decision. We follow Haurin, Hendershott and Wachter (1997) and
define the Down Payment Constraint to equal one if the household’s wealth is less than
10% of its desired house value. Desired house value is a predicted value, derived from a
sample-selection corrected estimation of homeowners’ house value on the demographic
and economic characteristics of households.
18
The t-statistics of the inverse Mill’s ratios in the sample of renters are: math (2.5),
reading (2.1) and behavior problems (4.5). In the sample of owners, the t-statistics are:
math (0.0), reading (1.5) and behavior problems (2.5). These results indicate significant
correlation of the error term in the tenure choice equation with the error terms in four
of the six child outcome equations.
Does Homeownership Affect Child Outcomes? 653
The impact of homeownership on the home environment is estimated in (4),
where the dependent variables are the two HOME scales. Estimation results
are listed in Table 2. The first data column reports the results for the estimation
of the cognitive stimulation/physical environment home scale. Significant variables with positive impacts include homeownership, mother’s AFQT, mother’s
education, mother’s age (with a declining marginal impact) and the church
attendance variables. Significant variables with negative impacts include the
child’s gender being male, mother’s race being black, the number of siblings
and the locality’s percentage of households in poverty.
Being a homeowner raises the value of the cognitive/physical environment by
10.7 points, with the mean of HOME-C being 47. Thus, compared with being a
renter, the measure of the cognitive/physical home environment of homeowners
is 23% higher. This dramatic increase occurs in spite of a large number of
controls for social, demographic and economic variables.19
We note that there is significant correlation from survey to survey of the
household-specific errors in the HOME-C equation, the correlation coefficient
being 0.42. In this and all following models, the hypothesis that household- or
child-specific errors are independent over time is firmly rejected. A test of the
hypothesis that there is a single common error term rather than household- or
child-specific errors also is rejected.
In the second data column, we show the impact of explanatory variables on
the level of emotional support in the home environment. Significant variables
with positive impacts include homeownership, mother’s age (declining marginal
impact), father’s age and mother’s education. Significant variables with negative
impacts include black, number of siblings, and the mother’s marital history
being single, remarried or becoming divorced/separated/widowed compared
with being continuously married. The negative effects of ending a marriage or
remarrying upon the measure of the emotional support in the home are large.
Being a homeowner raises the value of emotional support in the home environment by 5.9 points with the mean of HOME-E being 47. Compared with being
a renter, the measure of the level of emotional support in the home environment
of homeowners is 13% higher, ceteris paribus.
19
In the home environment and the child outcome equations, we tested for a nonlinear
impact of wealth. For example, we separated the wealth variable into a series of variables
where the first is zero if wealth is positive and equals the actual amount of wealth if
wealth is negative, the second is zero except for wealth equaling $0 to $5,000, the third
is zero except for wealth equaling $5,001 to $25,000, and so forth. Again we found no
impact for any of the wealth variables.
HOME-C
HOME-E
HOME-C
HOME-E
Variables
Coeff
t-stat
Coeff
t-stat
Variables
Coeff
t-stat
Coeff
t-stat
Homeownera
Male
Black
Mexican Hispanic
Other Hispanic
Health Limit
Siblings
Mother’s Age
Mother’s Age sq.
Mother’s HGC
Mother’s Mastery
Mother’s AFQT
Mother’s Wage
Father’s Wage
Father’s Age
Father’s HBC
Wealth
Nonlabor Income
10.67
−7.37
−4.21
−3.42
0.35
2.09
−2.41
3.86
−0.06
1.93
−0.13
0.22
0.27
0.08
0.01
0.24
−0.003
0.001
4.7
6.0
2.3
1.0
0.1
1.1
4.9
2.2
2.3
5.3
0.2
6.5
1.5
0.5
0.1
1.7
0.6
0.4
5.89
−1.82
−9.78
−3.32
7.16
−2.41
−1.28
4.81
−0.08
1.05
0.55
0.03
−0.01
0.11
0.19
0.29
−0.001
−0.004
2.5
1.6
5.7
1.0
1.5
1.2
2.7
2.6
2.6
3.1
0.9
0.9
0.0
0.6
2.5
1.9
0.3
0.9
MOM-ALONE
MOM-PAIR
MOM-OTHER
Church Attend-Low
Church Attend-Some
Church Attend-Often
Church Attend-High
SINGLE
GET MARRIED
REMARRY
MARITAL BREAKUP
% Black
% Hispanic
% Poverty
% High School Educ.
% College Educ.
Unemployment Rate
Median Income
Population Density
Crime Rate
Constant
2.68
1.28
−3.91
2.97
4.19
6.41
5.65
2.34
0.23
−1.64
0.19
0.11
0.02
−0.65
−0.08
−0.20
−0.20
0.11
0.01
0.04
−33.52
1.7
0.6
1.8
1.5
2.1
3.2
2.4
1.0
0.1
1.0
0.1
1.5
0.3
2.4
0.7
1.5
1.1
0.4
0.7
1.5
1.2
0.13
−1.24
1.46
−0.70
−0.52
−0.96
−2.59
−7.34
−2.83
−8.75
−13.12
0.04
0.07
−0.16
0.20
0.09
−0.35
0.01
−0.01
−0.03
−39.20
0.0
0.6
0.7
0.4
0.3
0.5
1.2
3.1
1.5
5.5
5.8
0.6
1.3
0.6
1.7
0.7
1.9
0.0
1.2
1.0
1.3
b
R-squared
a
b
0.42
0.25
0.31
0.27
The homeowner variable is the instrument based on the estimation in the Appendix.
is the autocorrelation of errors for a household, this due to the common error component; α H i .
654 Haurin, Parcel and Haurin
Table 2 Estimation results for the impact of homeownership on the home environment: 1988–1994.
Does Homeownership Affect Child Outcomes? 655
We conclude that homeownership impacts the levels of the cognitive stimulation/physical environment and emotional support environment of the home
in which a child lives. This result corresponds with our theory and is quite
plausible.
Equations (5) and (6) are the basis for the test for the impact of homeownership
on child outcomes. We use the instrument for homeownership derived from the
estimation in the Appendix to address the sample-selection issue. The estimation technique for (5) and (6) is a random effects panel data model. In the first
data column of Table 3, the dependent variable is the measure of a child’s mathematical achievement (Piat-Math). Significant explanatory variables (5% level)
with positive coefficients include the cognitive/physical and emotional support
home environment scales, mother’s achievement test score (AFQT), mother’s
and father’s education, a frequent or high level of church attendance, neighborhood median income and the neighborhood poverty rate (unexpected sign).20
The homeownership variable has a positive coefficient, and the coefficient is
significant at the 10% level. Significant variables with negative coefficients include low birth weight and more siblings. The period to period correlation of
child-specific estimation errors is positive and high: 0.50.
Accepting the point estimate of the homeownership variable implies that being a
homeowner directly raises PIAT-Math by 3.4 points, representing a 7% increase.
Further, being a homeowner raises the value of HOME-C by 10.7 points and
HOME-E by 5.9 points. The calculated indirect impact of homeowning on
PIAT-Math through an improved home environment is 0.8 points. Combined,
the total impact of homeownership on a child’s mathematical cognitive outcome
is to raise it about 9% compared to a family that rents, holding constant a large
number of social, demographic and economic variables.
The second set of results is for the measure of a child’s reading recognition.
Significant explanatory variables with positive coefficients include HOME-C,
HOME-E, mother’s AFQT, mother’s mastery and a high level of church attendance. Negative and significant effects occur for male children, more siblings
and a high local unemployment rate. The homeownership indicator has a positive coefficient, and it is significant at the 10% level.
Using the point estimate, being a homeowner raises PIAT-Reading directly by
3.2 points and indirectly by 0.7 points. Compared with an identical household
20
The measures of the community’s attributes are highly correlated; thus, it is difficult
to identify separate impacts. Also, as noted previously, they are relatively poor measures
of neighborhood characteristics.
656 Haurin, Parcel and Haurin
Table 3 Estimation results for a child’s cognitive and behavioral outcomes.
PIAT-Math
PIAT-Reading
BPI
Variables
Coeff
t-stat
Coeff
t-stat
Coeff
t-stat
Constant
Homeowner a
HOME-Cognitive/Physical
HOME-Emotional
Male
Black
Mexican Hispanic
Other Hispanic
Low Birth Weight
Health Limit
Siblings
Mother’s Age
Mother’s Age sq.
Mother’s HGC
Mother’s Mastery
Mother’s AFQT
Mother’s Wage
Father’s Wage
TOTAL HOURS MOM
WORK-YRS. 1–3
Father’s Age
Father’s HGC
Wealth
Nonlabor Income
MOM-ALONE
MOM-PAIR
MOM-OTHER
Church Attend-Low
Church Attend-Some
Church Attend-Often
Church Attend-High
SINGLE
GET MARRIED
REMARRY
MARITAL BREAKUP
% Black
% Hispanic
% Poverty
% High School Educated
% College Educated
Unemployment Rate
Median Income
Population Density
Crime Rate
(autocorrelation)
R-Squared
12.53
3.44
0.06
0.03
−0.60
−1.42
−2.06
−0.13
−4.72
−2.20
−1.13
−0.40
0.01
0.68
0.32
0.32
−0.03
0.10
0.03
0.5
1.7
4.4
2.5
0.5
0.8
0.6
0.0
2.0
1.4
2.5
0.3
0.2
2.0
0.5
10.0
0.2
0.7
0.1
55.10
3.15
0.04
0.04
−7.50
−0.78
−1.47
3.96
−4.28
−0.22
−1.63
−0.53
0.00
0.16
1.65
0.38
0.24
0.00
0.40
2.5
1.7
3.4
2.9
5.5
0.4
0.4
0.7
1.6
0.1
3.3
0.4
0.1
0.4
2.2
10.5
1.6
0.0
1.1
66.67
−1.69
−0.03
−0.04
5.18
−1.63
−2.73
0.10
2.51
3.85
0.23
0.64
−0.00
−0.67
−1.71
0.01
−0.28
0.07
−0.04
2.7
0.8
2.2
3.4
3.9
0.9
0.7
0.0
0.9
2.3
0.4
0.4
0.2
1.8
2.4
0.4
1.7
0.5
0.1
−0.12
0.25
0.006
0.005
−0.19
2.16
1.80
0.79
−0.86
3.76
5.88
0.12
−1.85
1.13
0.45
−0.13
−0.07
0.52
−0.03
−0.23
−0.18
0.84
0.01
0.02
0.50
0.23
1.9
2.0
1.3
1.5
0.1
1.1
0.9
0.4
0.5
2.0
2.6
0.1
1.1
0.8
0.2
1.9
1.3
2.1
0.3
1.9
1.2
3.1
1.0
1.0
−0.07
0.17
0.005
−0.001
0.46
−0.12
1.49
0.01
−0.93
3.10
5.87
3.28
−2.14
2.73
−0.01
0.03
0.07
−0.26
0.07
−0.10
−0.33
0.04
0.00
0.01
0.65
0.23
1.1
1.4
1.1
0.4
0.3
0.1
0.6
0.0
0.4
1.4
2.2
1.5
1.2
1.8
0.0
0.4
1.2
1.0
0.6
0.8
2.2
0.1
0.4
0.2
0.01
−0.06
0.001
−0.001
2.37
4.02
3.94
0.51
−1.44
1.15
2.46
2.19
0.65
3.30
2.05
0.10
0.03
−0.14
0.04
−0.04
−0.05
−0.23
−0.01
−0.02
0.58
0.06
0.1
0.5
0.2
0.3
1.4
1.8
1.7
0.2
0.7
0.5
1.0
0.9
0.3
2.1
1.0
1.4
0.5
0.5
0.3
0.3
0.3
0.8
0.9
0.8
a
Instrumental variable.
Does Homeownership Affect Child Outcomes? 657
that rents, these results indicate that residence in an owned home raises a child’s
reading score by about 7%.
The final column reports estimation results for the index of a child’s behavior
problems (BPI). Recalling that the index is higher when behavior problems
are greater, the expected coefficient signs are the opposite of those for the
models of cognition. Negative and significant coefficients occur for HOME-C,
HOME-E and mother’s mastery. Significant and positive coefficients occur for
male children, children with health limitations and if the mother divorces and
remarries. The homeownership indicator has the expected negative coefficient
but the t-statistic is only 0.8.
Again based on the point estimate, compared with a similar renter, homeowning
directly reduces the measure of the child’s behavior problems by 1.7 points,
equaling 2.6% of the mean value of BPI. Homeownership also changes the
cognitive/physical and emotional support home environments, further reducing
the BPI by 0.9 points. The cumulative impact is that homeownership reduces
the index of child behavior problems by about 3%, but the lack of statistical
significance suggests that the impact could be only 1%.
We next turn to the question of whether the duration of homeowning matters.
We reestimated the system of equations substituting time spent as an owner
for the homeownership indicator variable. In Table 4, we report only the key
coefficients for the ownership and home environment variables. We find that
use of the duration of homeowning produces a series of results that support our
hypotheses that homeownership increases a resident child’s cognitive outcomes
and decreases behavioral problems. The impact is both direct and through improvements to the cognitive stimulation/physical environment and emotional
support levels in the home.21
Estimation methods make a difference. If we ignore the panel nature of the
data and estimate the equations with simple OLS, then the results for the key
variables are quite different. Table 5 lists the coefficients and t-statistics for
the five equations. In our sample, OLS results greatly overstate the impact
of homeowning both on the home environment and direct effects on child
outcomes. Instead of math, reading and behavioral improvements of 7, 6 and
4% for homeowners, the estimated changes using OLS are 21, 14 and 17%.22
21
If included together in the estimation, the homeownership indicator and the duration
of homeowning variable take opposite signs due to their high correlation.
22
We also separated the sample into black and nonblack subsamples. The estimation
results in the black subsample for the two home environment equations are statistically
658 Haurin, Parcel and Haurin
Table 4 Estimation results using the duration of homeowning as the explanatory
variable.
HOME-C
Variable
HOME-E
PIAT-Math
PIAT-Reading BPI
Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat
Duration
1.40
of Owner
HOME-C
—
HOME-E
—
4.6
0.83
2.6
—
—
Coeff
t-stat
0.49
1.8
0.65
2.6
−0.39 1.4
0.06
0.03
4.4
2.5
0.04
0.03
3.4
2.9
−0.03 2.1
−0.04 3.4
Table 5 Estimation results ignoring panel nature of the data.
HOME-C
Variable
HOME-E
PIAT-Math
PIAT-Reading BPI
Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat
Duration
19.01 8.8
of Owner
HOME-C
—
HOME-E
—
8.06
—
—
3.7
Coeff
t-stat
7.02
3.5
4.19
2.0
−8.99 4.2
0.11
0.05
7.7
3.2
0.14
0.07
8.9
4.4
−0.08 4.9
−0.09 6.0
Thus, accounting for household- and child-specific effects matters substantially
in our analysis.
Recognition of the possible sample-selection bias led to our use of an instrumental variable for homeownership. This correction has some impact on the
results. We find that the coefficients of the homeownership indicator are smaller
in the cognition equations if we do not use an instrument, but the coefficient
of homeowning in the BPI equation is unexpectedly positive and significant.
After correcting for selection bias (which is largest in the BPI equation; see
footnote 15), the instrumental variable approach yields the expected direction
of impact.
Understanding the impact of control variables on child outcomes is important
as well. Combining the direct and indirect effects, we find that male and black
children score lower on math and reading than females and whites, but their
significant and very similar to those for nonblack households. The results in the black
subsample for child cognition and behavior problems are similar to those for nonblack
households (all coefficients have their expected signs), but t-statistics average only 1.2.
The Hispanic subsample is too small to separately analyze.
Does Homeownership Affect Child Outcomes? 659
reported behavioral problems are greater. Low birth weight has a long-term negative effect on math achievement, and children with current health limitations
have more behavior problems. Children with more siblings score lower on math
and reading, a result consistent with the family-resource-dilution hypothesis.
Children of older mothers and with more highly educated parents score higher
on math and somewhat higher on reading and have fewer behavioral problems.
Higher wages, wealth and nonlabor income do not have a statistically significant impact on child outcomes. The greater the mother’s aptitude, the greater
her children’s math and reading test scores. The greater her perceived control of
life, the higher her children’s reading scores and the lower the behavioral problems index. The living arrangement in which the mother was reared has little
impact nor does the amount she worked during the child’s first three years of
life. Children of parents who report relatively high levels of church attendance
score higher on cognitive tests. Relative to mothers who remained married from
the child’s birth to the survey date, the children of mothers who remained single, separated and then remarried, or became divorced/separated/widowed have
lower scores on the emotional support home environment scale and have more
behavioral problems.
Conclusion
In the United States, homeowners are subsidized through the tax codes; the
amount of the subsidy is estimated to exceed $60 billion annually. In addition,
many programs sponsored by the Department of Housing and Urban Development and state and local governments encourage homeownership. Although
increasing the homeownership rate is a goal of the federal and state governments, relatively little is known about the impact of homeowning on the resident
households. We add to the literature about the consequences of homeowning by
studying its impact on the cognitive and behavioral outcomes of a household’s
children.
The existing literature about child outcomes suggests a large vector of control variables is needed to isolate the impact of homeownership status. Further
complicating the analysis of child outcomes is the problem that unobserved
parental characteristics might lead to sorting by residential tenure status and
cause sample-selection bias if OLS is used for the analysis. Addressing this
selection problem is important to the correct isolation of the impact of homeownership on child outcomes.
We use four waves of a national data set to analyze the relationship of owning
a home to three child outcomes including math and reading cognition and a
measure of behavior problems. The wide scope of the data in the survey allows
660 Haurin, Parcel and Haurin
us to include as control variables many social, demographic and economic
variables suggested by previous studies of child outcomes. Using panel data
allows us to control for household- and child-specific unobserved factors, an
approach infrequently implemented in prior studies, but which proves to be
very important. We also use the method of instrumental variables to address the
problem of sample-selection bias.
Our results suggest that owning a home leads the resident household to invest
in their property and to produce a higher quality home environment. We also
find that a child’s cognitive outcomes are up to 9% higher in math achievement
and 7% higher in reading achievement for children living in owned homes. An
index of children’s behavior problems is up to 3% lower if the child resides
in an owned home. These results occur even when we control for numerous
parental economic, demographic and social characteristics. We also control for
the child’s gender and health, number of siblings and nine characteristics of
the household’s locality. Thus, in a well-controlled study, accounting for unobserved child characteristics, and accounting for unobserved parental characteristics that might lead to a spurious correlation between homeowning and child
outcomes, we find substantial support for the hypothesis that homeownership
increases child cognition and reduces behavior problems.
Existing housing policies encouraging homeownership are sometimes targeted
at households who would become first-time homeowners. Our finding that
homeownership enhances child outcomes suggests that housing policies should
be targeted at rental households that have children. Not all renters should become homeowners, but if homeownership conveys benefits to children, it is
sensible for public policies to account for this result. Quickening the transition
to owning of families with children would expose these children to better and
more stable home environments for a longer period.
There continues to be illegal discrimination in the housing market, particularly
regarding aspects of the homeownership decision. Our study supports the conclusion that any reduction in homeownership due to illegal discrimination also
has the effect of reducing the level of cognition and increasing the behavioral
problems of the children of households that are the targets of discrimination.
Reducing illegal discrimination may not only help solve the problem of spatial mismatch of jobs and residences, but such reduction may also result in
long-term gains of the children in these households.
There is continuous interest in improving the educational outcomes in K–12
public schools. Policies proposed and often implemented in support of improving K–12 outcomes include increased funding for schools or offering parents
Does Homeownership Affect Child Outcomes? 661
alternatives to public schools such as charter schools and school vouchers. However, for decades, research has found mixed or no impact of these policies on
school outcomes (Hanushek 1986, 1996). Rather, parental and home environment influences are the dominant factors. Our results support a new approach
to the problem of achievement levels in K–12 education. Improving the home
environment can be achieved by increasing the homeownership rates of parents.
Our study’s results correspond with the dominant findings in the education literature, specifically, that improving the home environment leads to better child
cognition and better child social adjustment.
We thank David Brasington, Nam-yll Kim, Donghui Qiu, Mikaela Dufur and Robert
Dietz for their assistance. We thank the participants of the Harvard Joint Center for
Housing Studies Low Income Homeownership Symposium for comments.
References
Aaronson, D. 2000. A Note on the Benefits of Homeownership. Journal of Urban
Economics 47: 356–369.
Achenbach, T.S. and C. Edelbrock. 1981. Behavioral Problems and Competencies Reported by Parents of Normal and Disturbed Children Aged Four through Sixteen. Monographs of the Society for Research in Child Development Serial No. 188, 46(1).
Achenbach, T.S. and C. Edelbrock. 1983. Manual for the Child Behavior Checklist
and Revised Child Behavior Profile. Department of Psychiatry, University of Vermont:
Burlington, VT.
Achenbach, T.S., S.H. McConaughy and C.T. Howell. 1987. Child/Adolescent Behavioral and Emotional Problems: Implications of Cross-Informant Correlations for Situational Specificity. Psychological Bulletin 101: 213–232.
American Chamber of Commerce Research Association. 1993. Cost of Living Index
Manual. Louisville Chamber of Commerce: Louisville, KY.
Austin, D.M. and Y. Baba. 1990. Social Determinants of Neighborhood Attachments.
Sociological Spectrum 10: 59–78.
Baker, P.C., C. Keck, F.L. Mott and S. Quinlan. 1993. NLSY Child Handbook Revised Edition. Center for Human Resource Research, Ohio State University: Columbus,
OH.
Balfor, D. and J. Smith. 1996. Transforming Lease-Purchase Housing Programs for Low
Income Families: Towards Empowerment and Engagement. Journal of Urban Affairs
18: 173–188.
Barnow, B.S., G.G. Cain and A.S. Goldberger. 1981. Issues in the Analysis of Selectivity
Bias. Evaluation Studies Review Annual. W.E. Stromsdorfer and G. Farkas, editors. Sage:
Beverly Hills, CA.
Baum, T. and P. Kingston. 1984. Homeownership and Social Attachment. Sociological
Perspectives 27: 159–180.
Becker, G. 1965. A Theory of the Allocation of Time. Economic Journal 75: 493–
517.
Blake, J. 1989. Family Size and Achievement. University of California Press: Berkeley,
CA.
662 Haurin, Parcel and Haurin
Boehm, T.P. and A.M. Schlottmann. 1999. Does Homeownership by Parents Have
an Economic Impact on Their Children? Journal of Housing Economics 8: 217–
232.
Bradley, R.H. and B.M. Caldwell. 1981. The HOME Inventory: A Validation of the
Pre-School Scale for Black Children. Child Development 52: 708–710.
Bradley, R.H. and B.M. Caldwell. 1984a.The HOME Inventory and Family Demographics. Developmental Psychology 20: 315–320.
Bradley, R.H. and B.M. Caldwell. 1984b. The Relation of Infants’ Home Environments
to Achievement Test Performance in First Grade: A Follow-Up Study. Child Development 55: 803–809.
Bradley, R.H., B.M. Caldwell, S.L. Rock, H.M. Hamrick and P. Harris. 1988. Home
Observation for Measurement of the Environment: Development of a Home Inventory
for Use with Families Having Children 6 to 10 Years Old. Contemporary Educational
Psychology 13: 58–71.
Briggs, X.d.S. 1998. Brown Kids in White Suburbs: Housing Mobility and the Many
Faces of Social Capital. Housing Policy Debate 9: 177–221.
Case, A.C. and L.F. Katz. 1991. The Company You Keep: The Effects of Family and
Neighborhood on Disadvantaged Youths. Working paper No. 3705. National Bureau of
Economic Research: Cambridge, MA.
Caspi, A., G.H. Elder, Jr. and D.J. Bern. 1987. Moving Against the World: Life-Course
Patterns of Explosive Children. Developmental Psychology 23: 308–313.
Center for Human Resource Research. 1994. NLSY Child Handbook. Author: Columbus,
OH.
Clark, R. 1992. Neighborhood Effects on Dropping Out of School among Teenage Boys.
Working paper. Urban Institute: Washington, DC.
Coleman, J.S. 1988. Social Capital in the Creation of Human Capital. American Journal
of Sociology 94: S95–S120.
Coleman, J.S. 1990. Foundations of Social Theory. Harvard University Press:
Cambridge, MA.
Cox, K. 1982. Housing Tenure and Neighborhood Activism. Urban Affairs Quarterly
18: 107–129.
Crane, J. 1991. The Epidemic Theory of Ghettos and Neighborhood Effects on Dropping
Out and Teenage Childbearing. American Journal of Sociology 96: 1126–1159.
Cronbach, L.J. 1951. Coefficient Alpha and the Internal Structure of Tests. Psychometrika 16: 297–334.
Currie, J. and A. Yelowitz. 2000. Are Public Housing Projects Good for Kids? Journal
of Public Economics 75: 99–124.
Dietz, R.D. and D.R. Haurin. 2002. The Social and Private Consequences of Homeownership. Working paper. Ohio State University: Columbus, OH.
DiPasquale, D. and E. Glaeser. 1999. Incentives and Social Capital: Are Homeowners
Better Citizens? Journal of Urban Economics 45: 354–384.
Dornbusch, S.M., P.L. Ritter and L. Steinberg. 1991. Community Influences on the
Relation of Family Statuses to Adolescent School Performance: Differences between
African Americans and Non-Hispanic Whites. American Journal of Education 100:
543–567.
Downey, D. 1995. When Bigger is not Better: Number of Siblings, Parental Resources,
and Educational Performance. American Sociological Review 60: 746–761.
Duncan, G.J. 1994. Families and Neighbors as Sources of Disadvantage in the Schooling
Decisions of White and Black Adolescents. American Journal of Education 103: 20–
53.
Does Homeownership Affect Child Outcomes? 663
Elardo, R.D. and R.H. Bradley. 1981. The Home Observation for Measurement of the
Environment (HOME) Scale: A Review of Research. Developmental Review 1: 113–
145.
Elton, P.J. and J.M. Packer. 1986. A Prospective Randomised Trial of the Value of ReHousing on the Grounds of Mental Ill-Health. Journal of Chronic Disease 39: 221–
227.
Engelhardt, G. 1994. House Prices and the Decision to Save for Down Payments. Journal
of Urban Economics 36: 209–237.
Evans, G.W., N.M. Wells, H.Y. Chan and H. Saltzman. 2000. Housing Quality and
Mental Health. Journal of Consulting and Clinical Psychology 3: 526–530.
Evans, W., W. Oates and R. Schwab. 1992. Measuring Peer Group Effects: A Study of
Teenage Behavior. Journal of Political Economy 100: 966–991.
Executive Office of the President. 2000. The Fiscal Year 2001 Budget of the United
States. Washington, DC.
Flavin, M. and T.Y. Yamashita. 2000. Owner Occupied Housing and the Composition
of the Household Portfolio. Working paper. University of California at San Diego: San
Diego, CA.
Follain, J.R. and L.S. Melamed. 1998. The False Messiah of Tax Policy: What Elimination of the Home Mortgage Interest Deduction Promises and a Careful Look at What it
Delivers. Journal of Housing Research 9: 179–199.
Forgatch, M.S., G.R. Patterson and M.L. Skinner. 1988. A Mediational Model for the
Effect of Divorce on Antisocial Behavior in Boys. Impact of Divorce, Single Parenting,
and Step-Parenting on Children. E.M. Hetherington and J.D. Arestah, editors. Erlbaum:
Hillsdale, NJ.
Galster, G. 1983. Empirical Evidence on Cross-Tenure Differences in Home Maintenance and Conditions. Land Economics 59:107–113.
Galster, G. 1987. Homeowners and Neighborhood Reinvestment. Duke University Press:
Durham, NC.
Gatzlaff, D.H., R.K. Green and D.L. Ling. 1998. Revisiting Cross-Tenure Differences
in Housing Maintenance. Land Economics 74: 328–342.
Green, R. and M. White. 1997. Measuring the Benefits of Homeowning: Effects on
Children. Journal of Urban Economics 41: 441–461.
Greene, W. 1993. Econometric Analysis. Second edition. Macmillan Publishing Co.:
New York, NY.
Greene, W. 1995. LIMDEP, Version 7.0 User’s Manual. Econometric Software Inc.:
Bellport, NY.
Hanushek, E.A. 1986. The Economics of Schooling: Production and Efficiency in Public
Schools. Journal of Economic Literature 24: 1141–1177.
Hanushek, E.A. 1996. School Resources and Student Performance. Does Money Matter?
The Effect of School Resources on Student Achievement and Adult Success. G. Burtless,
editor. Brookings Institution: Washington, DC.
Hanushek, E.A., J. Kain and S. Rivkin. 1999. The Costs of Switching Schools. Working
paper. Rochester University: Rochester, NY.
Haurin, D.R., R.D. Dietz and B.A. Weinberg. 2002. The Impact of Neighborhood Homeownership Rates: A Review of the Theoretical and Empirical Literature. Working paper.
Ohio State University: Columbus, OH.
Haurin, D.R., P.H. Hendershott and D. Kim. 1994. Housing Decisions of American
Youth. Journal of Urban Economics 35: 28–45.
Haurin, D.R., P.H. Hendershott and D. Ling. 1988. Homeownership Rates of Married
Couples: An Econometric Investigation. Housing Finance Review 7: 85–108.
664 Haurin, Parcel and Haurin
Haurin, D.R., P.H. Hendershott and S. Wachter. 1996. Wealth Accumulation and Housing Choices of Young Households: An Exploratory Investigation. Journal of Housing
Research 7: 33–57.
Haurin, D.R., P.H. Hendershott and S. Wachter. 1997. Borrowing Constraints and
the Tenure Choice of American Youth. Journal of Housing Research 8: 137–
154.
Haurin, R.J. 1992. Patterns of Childhood Residence and the Relationship to Young Adult
Outcomes. Journal of Marriage and the Family 54: 846–860.
Haurin, R.J. and F. Mott. 1990. Adolescent Sexual Activity in the Family Context: The
Impact of Older Siblings. Demography 27: 537–557.
Haveman, R. and B. Wolfe. 1995. The Determinants of Children’s Attainments:
A Review of Methods and Findings. Journal of Economic Literature 33: 1829–
1878.
Heckman, J. 1979. Sample Selection Bias as a Specification Error. Econometrica 47:
153–161.
Hendershot, P.H. and J. Shilling. 1982. The Economics of Tenure Choice, 1955–79.
Research in Real Estate. C.F. Sirmans, editor. JAI Press: Greenwich, CT.
Hunter, A. 1975. The Loss of Community: An Empirical Test Through Replication.
American Sociological Review 40: 537–551.
Ineichen, B. and D. Hooper. 1974. Mental Health and the Built Environment. Taylor and
Francis: London.
Jeffers, L. and J. Dobos. 1984. Communication and Neighborhood Mobilization. Urban
Affairs Quarterly 20: 97–112.
Kohlberg, L., J. LaCrosse and D. Ricks. 1972. The Predictability of Adult Mental Health
from Childhood Behavior. Manual of Child Psychopathology. B.B. Wolman, editor.
McGraw-Hill: New York, NY.
Lead-Based Paint Hazard Reduction and Financing Task Force. 1995. Putting the Pieces
Together: Controlling Lead Hazards in the Nation’s Housing. Washington, DC.
Lee, B.A., R.S. Oropesa and J. Kanan. 1994. Neighborhood Context and Residential
Mobility. Demography 31: 249–270.
Mayer, N. 1981. Rehabilitation Decisions in Rental Housing. Journal of Urban
Economics 10: 76–94.
McCarthy, G., S. Van Zandt and W. Rohe. 2001. The Economic Benefit and Costs of
Homeownership: A Critical Assessment of the Literature. Working paper No. 01–02.
Research Institute for Housing America: Washington, DC.
Mechanic, D. 1980. Mental Health and Social Policy. Second edition. Prentice-Hall:
Englewood Cliffs, NJ.
Menaghan, E.G. and T.L. Parcel. 1991. Determining Children’s Home Environments:
The Impact of Maternal Characteristics and Current Occupational and Family Conditions. Journal of Marriage and the Family 53: 417–431.
Menaghan, E.G. and T.L. Parcel. 1995. Social Sources of Change in Children’s Home
Environments: The Effects of Parental Occupational Experiences and Family Conditions. Journal of Marriage and the Family 57: 1–16.
Mott, F. 1991. Developmental Effects of Infant Care: The Mediating Role of Gender
and Health. Journal of Social Issues 47: 139–158.
Newman, S., J. Harkness and W.J. Yeung. 1998. Neighborhood Poverty, Assisted
Housing, and the Educational Attainment of Children. Working paper. Johns Hopkins
University: Baltimore, MD.
Office of Federal Housing Enterprise Oversight. 1998. House Price Index. Washington,
DC.
Does Homeownership Affect Child Outcomes? 665
Parcel, T.L. and E.G. Menaghan. 1990. Maternal Working Conditions and Child Verbal
Facility: Studying the Intergenerational Transmission of Inequality from Mothers to
Young Children. Social Psychology Quarterly 53: 132–147.
Parcel, T.L. and E.G. Menaghan. 1994a. Early Parental Work, Family Social Capital and
Early Childhood Outcomes. American Journal of Sociology 99: 972–1009.
Parcel, T.L. and E.G. Menaghan. 1994b. Parents’ Jobs and Children’s Lives. Aldine De
Gruyter: New York, NY.
Rogers, S., T.L. Parcel and E.G. Menaghan. 1991. The Effects of Maternal Working Conditions and Mastery on Child Behavior Problems: Studying the Intergenerational Transmission of Social Control. Journal of Health and Social Behavior 32: 145–
164.
Rohe, W. and V. Basolo. 1997. Long-Term Effects of Homeownership on the SelfPerceptions and Social Interactions of Low-Income Persons. Environment and Behavior
29: 793–819.
Rohe, W., G. McCarthy and S. Van Zandt. 2000. The Social Benefits and Costs of
Homeownership: A Critical Assessment of the Research. Working paper No. 00-01.
Research Institute for Housing America: Washington, DC.
Rohe, W. and M. Stegman. 1994a. The Impacts of Homeownership on the Self Esteem,
Perceived Control and Life Satisfaction of Low Income People. Journal of the American
Planning Association 60: 173–184.
Rohe, W. and M. Stegman. 1994b. The Impact of Homeownership on the Social
and Political Involvement of Low-Income People. Urban Affairs Quarterly 30: 152–
172.
Rohe, W. and L. Stewart. 1996. Homeownership and Neighborhood Stability. Housing
Policy Debates 7: 37–81.
Rossi, P.H. and E. Weber. 1996. The Social Benefits of Homeownership: Empirical
Evidence from National Surveys. Housing Policy Debates 7: 1–35.
Rotter, J.B. 1966. Generalized Expectancies for Internal vs. External Control of Reinforcements. Psychological Monographs 80: 1–28.
Saegert, S. 1982. Environment and Children’s Mental Health: Residential Density and
Low Income Children. Handbook of Psychology and Health. A. Baum and J.E. Singers,
editors. Erlbaum: Hillsdale, NJ.
Shilling, J.D., C.F. Sirmans and J.F. Dombrow. 1991. The Effect of Tenant Occupancy
on the Measurement of Depreciation in Housing. Journal of Housing Economics 1:
368–383.
Staiger, D. and J.H. Stock. 1997. Instrumental Variables Regression with Weak Instruments. Econometrica 65: 557–586.
Zill, N. 1988. Behavior, Achievement, andHealth Problems among Children in Stepfamilies: Findings from a National Survey of Child Health. The Impact of Divorce, Single
Parenting, and Stepparenting on Children. E.M. Hetherington and J.D. Arasteh, editors.
Erlbaum: Hillsdale, NJ.
Appendix
Table 6 give an estimation of the tendency to own a home using a random
efffects probit model and the duration of homeownership using a tobit model.
666 Haurin, Parcel and Haurin
Table 6 Estimation of the tendency to own a home using a random effects probit
model and the duration of homeownership using a Tobit model.a
Homeownership
Duration of Owning
Variables
Coeff.
t-statistic
Coeff.
t-statistic
Constant
House Price
Down Payment
Constrained
Mother’s Wage
Father’s Wage
Wealth
Current Interest Rate
Lagged Interest rate
Live in MSA
Population Density
Child Age
Child Age-squared b
Black
Mexican Hispanic
Other Hispanic
Mother’s Age
Father’s Age
# Siblings
(autocorrelation)
−5.32
−0.73
−1.44
2.8
3.4
15.8
−24.19
−1.25
−6.28
6.9
3.3
23.8
−0.04
0.08
0.001
0.04
−0.05
−0.31
−0.01
0.03
−0.01
−1.12
0.39
−0.01
0.19
0.03
−0.12
0.87
1.8
4.1
2.7
0.3
0.2
2.1
3.0
2.9
3.9
6.5
0.9
0.0
5.4
3.3
2.2
21.1
−0.08
0.15
0.000
0.02
0.44
−0.26
−0.02
0.04
−0.02
−3.14
0.49
−0.24
0.07
0.09
−0.41
1.9
4.7
0.3
0.1
0.9
1.0
4.4
1.2
1.4
11.8
0.9
0.2
13.5
7.0
4.4
a
In the sample, 40% of households are homeowners. Of the 2,470 households who
rent, the estimation predicts tenure correctly in 86% of the observations. Of the 1,634
households who own, the equation predicts tenure correctly in 68% of the cases.
b
Child Age-squared is measured in hundreds.