House Price Volatility and
the Housing Ladder
JAMES P. SMITH
This paper series made possible by the NIA funded RAND Center for the Study
of Aging (P30AG012815) and the NICHD funded RAND Population Research
P A P E R
This product is part of the RAND
Labor and Population working
paper series. RAND working papers
are intended to share researchers’
latest findings and to solicit informal
peer review. They have been approved
for circulation by RAND Labor and
Population but have not been
formally edited or peer reviewed.
Unless otherwise indicated, working
papers can be quoted and cited
without permission of the author,
provided the source is clearly referred
to as a working paper. RAND’s
publications do not necessarily
reflect the opinions of its research
clients and sponsors.
is a registered trademark.
House price volatility and the housing ladder
Institute for Fiscal Studies and University of Manchester
Institute for Fiscal Studies and University College London
Institute for Fiscal Studies
James P. Smith
This paper investigates the effects of housing price risk on housing choices over the life-cycle.
Housing price risk can be substantial but, unlike other risky assets which people can avoid, the
fact that most people will eventually own their home creates an insurance demand for housing
assets early in life. Our contribution is to focus on the importance of home ownership and housing
wealth as a hedge against future house price risk for individuals moving up the ladder—people
living in places with higher housing price risk should own their first home at a younger age,
should live in larger homes, and should be less likely to refinance. These predictions are tested
and shown to hold using panel data from the United States and Great Britain.
JEL: D12, D91
Acknowledgements: The authors gratefully acknowledge the financial support of the Economic
and Social Research Council through the research grant RES-000-22-0513. Blundell and Banks
would like to thank the ESRC Centre for the Microeconomic Analysis of Public Policy at IFS for
co-funding. James Smith’s research was supported by grants from the National Institute on Aging
and benefited from the expert programming assistance of Patty St. Clair and David Rumpel.
One of the most critical consumption and investment decisions that individuals
and families make over their life-cycle involves the amount of housing services to
consume and whether or not to combine consumption with ownership. Housing is an
important component of consumption, but not simply because it absorbs a large fraction
of the household budget—which it does. Where we live and how much we decide to
spend on housing is a key ingredient to the amenities and life-style we have chosen for
our families and ourselves. But housing, or more particularly housing wealth, can be even
more critical as an investment as it is typically by far the biggest marketable asset in the
household portfolio for most people.
The contribution of this paper is to bring together two key determinants of
housing consumption and home ownership decisions into an empirical model of housing
outcomes. The first of these is the housing ladder. Rather than modeling home ownership
as a one-time durable purchase, we model it as a series of purchase decisions, or a
housing ladder, where the desired flow of housing services rises with family formation
and growing family size over the life cycle. The second is the acknowledgement of the
role of future house price risk. In some geographic markets, housing can be a risky asset
with high levels of unpredictable price volatility while in other places the prospect of
capital gains or losses in housing are understandably not the subject of much social
Our contribution is to focus on the importance of ownership as a hedge against
future house price risk as individuals move up the ladder. We use a stylized model to
show that increasing house price risk acts as an incentive to become a home owner earlier
in the life-cycle and, once an owner, to move more rapidly up the housing ladder.
Increases in volatility are shown to increase ownership and to increase the quantity of
housing wealth conditional on ownership in earlier periods of the life-cycle. We then
establish that these relationships hold empirically using panel data on families in different
geographic markets in Britain and in the US.
Housing needs change over the life-cycle and the decision of when to buy the first
property and at what point to move up the ladder is a key life-cycle decision. For
example, Ortalo-Magne and Rady (2004, 2006) note the importance of new entrants at
the bottom of the ladder for the determination of housing transactions along the whole
ladder. Ermisch and Pevalin (2004) document the importance of childbearing and family
formation decisions on housing choices. We follow this lead by allowing the demand for
housing consumption and movements up the ladder to depend directly on the
demographic profile of the family. We then add to this the enhanced incentive to own and
to move up the ladder created by more volatile house prices.
The idea that home ownership can be seen as a hedge against uncertainty in the
price of housing services has many precedents. For example, Sinai and Souleles (2005)
use this observation to carefully show the increased demand for ownership when rental
price uncertainty is higher. Our contribution instead is to focus on the importance of
ownership and the quantity of housing owned as a hedge against future house price risk
as individuals move up the ladder. We examine the impact of volatility on both
ownership and on measures of the quantity of housing wealth conditional on ownership.
Both are shown to rise with increased house price volatility.
In contrast to other risky assets in which risk-averse individuals can simply
choose to avoid them, everyone must consume housing, and the vast majority of people
desire to and eventually do end up owning their own home. In addition, for most
individuals the demand for housing will rise over the life-cycle as family size increases.
The combination of these factors results in an insurance role for housing wealth in early
life that drives the predictions we investigate in our empirical analysis.
Using panel data from the UK and the US, we test the implications of the ladder
and price volatility on the decision on when to become a homeowner, how much housing
to consume, and whether to refinance out of housing equity. In the presence of volatility
in house prices, housing has three roles—investment, consumption, and insurance against
price fluctuations for future movements up the housing ladder. A simple theoretical
discussion illustrates these effects, and the predictions for home ownership, and housing
wealth accumulation are drawn out.
Because housing price volatility is spatially variable, we test the importance of the
role of volatility in housing decisions empirically using comparable panel data from the
US and the UK. There are significant differences in housing price variability between and
within these two countries. But in addition there are also differences in the tax treatment
of mortgage debt, the nature of mortgage arrangements, and even the level of geographic
mobility of younger households. Consequently a test relying on between country
differences is unlikely to isolate the effects of interest. In our analysis we show that,
while the international differences are indeed in accordance with the predictions of our
model, the model also performs well when estimated from within country variation in
each of the countries we consider, despite their rather wide institutional differences.
The analysis in this paper is in five sections. Section I documents a critical and
salient fact—a steep housing ladder with age which is coincident with changing
demographics over the life-cycle that are common across the two countries. Section II
shows the large special dispersion in house price volatility within and between the UK
and US. Section III then discusses the implications of housing price variability for
housing choices in a simple life-cycle framework. In Section IV the model predictions
concerning the age of initial home ownership, the decision to refinance, and the shape of
housing wealth and the number of rooms are put to the test. In the final section we
summarize our conclusions.
I. The Housing Ladder
Even without credit constraints or income uncertainty, individuals would not choose to
consume the same flow of housing services at all times in their lives. People may start by
moving out of the parental home into a small rented or purchased apartment or flat of
their own. When they marry, they may know that two may well live more cheaply than
one but they generally do not want to live in smaller places and often may want to own a
bigger but still modest first home. Children then appear on the scene and eventually will
age into rooms of their own—all of which requires a bigger if not better home.
A simple way of illustrating this point is to examine how the size of homes people
live in changes with age. Table 1 shows the age profile of mean number of rooms of
household heads for owners and renters alike in the US and UK using the Panel Study of
Income Dynamics (PSID) in the US and the British Household Panel Study (BHPS).1
Note that the number of rooms in the British data excludes kitchens and bathrooms and so
the number of rooms is not strictly comparable across the two countries. In both countries
there is a strong increase in size of house as the head of household grows older, flattening
out around the age 40 but rising steeply from the 20s to the 30s. The general shape of the
ladder is similar in the two countries.2 It is important to note that the steep part of the
ladder is not simply the consequence of changing tenure status from renter to owner.
While owned homes are always larger than rented ones on average, the steep early ladder
characterizes both rented and owned properties.3
Another way of seeing this transition is to examine the increase in home size at
the time of purchase among new and repeat buyers. This is shown in Table 2. New buyers
are defined as those who were previously renters in the prior wave of PSID or BHPS so
that especially at young ages this often will be their first owned home. Repeat buyers
were previously also homeowners so that this change now reflects changes in the size of
owner occupied housing. In the US, while the transition from renter to owner involves a
larger increment in house size, people are also clearly trading up in the early part of the
life cycle when they purchase their second and subsequent homes. This effect is even
stronger in the UK—on average first time buyers purchase houses that are bigger
comparable than their rented house, but bigger movements up the ladder, defined in terms
of increments to the number of rooms, tend to take place for repeat buyers.
We view the shape of the ladder as demographically determined as individuals
marry, form families with children growing, eventually complete their family building
with the by now older children leaving home to go off on their own. Figures 1a and 1b
plot the cumulative distribution of individuals who have completed their fertility by age.4
The steepness of this cumulative distribution mimics closely the overall shape of the
housing ladder—a steep incline during the 20s and 30s with a flattening out during the
40s. In fact, between ages 25 and the late 30s, this cumulative distribution of competed
fertility is almost linear, with each year of age increasing the fraction that has finished
childbearing by 5 percentage points. For example, around age 31, half of all American
individuals have completed their fertility with three out of every four doing so by age 36.
The shape, and level, of the profile corresponds extremely closely to that observed in the
UK over the same ages.
Children turning 5 years old may be at a critical stage for housing decisions since
parents may choose places to live with the quality of schools in mind and may want to
stay longer in the same place. This could be another indicator of reaching the top of the
housing ladder and arrival in the ‘family home.’ With this in mind, Figures 1a and 1b also
plot the cumulative fraction of individuals who ever had child at least 5 years old. Not
surprisingly, compared to the cumulative completed fertility, this figure is shifted out to
the right so that if age 5 is a useful marker, reaching the top of the ladder takes place for
the median family in the mid to late thirties. Nevertheless, as with the completed family
size profile, the proportion rises steeply over the life cycle up to age 40 in parallel to the
sharp rise in the number of rooms demonstrated over the same ages. Finally, Figures 1a
and 1b also plot the proportion with their own children aged 5 or over currently in the
household, as a measure of contemporaneous housing needs. Again the similarities
between US and UK are striking—in both countries after age 40 there is a sharp decline
in young children at home, an indication of an eventual demographic rationale for
downsizing in later life.
II House Price Volatility
Figure 2 shows real indices of country-wide average house prices for the US and
UK over the period 1974 to 1998 with both series normalized to take a value of one in
1980. Immediately apparent is the much larger volatility of housing prices in the UK,
with real prices rising by 50% over the period 1980 to 1989 and then falling back to its
previous value by 1992. Over the period as a whole, however, real returns were similar
across the two countries.
Although such difference will be instructive when looking at differences in
housing choices across the two countries, the majority of our testing will rely instead on
within-country differences in house price volatility in each of the two countries. The UK
and the US indexes both hide considerable differences across regions with some places
being much more volatile in housing prices than others. In Figures 3a and 3b we present
house prices from regional sub indices, grouped to show house price trends in the more
and less volatile areas.
The variation across American states in housing price volatility is large. Using the
standard deviation in real prices (relative to a 1980 base) as the index, Massachusetts
ranks at the top with price swings between peak and trough over this period of more than
two to one. At the other extreme lies South Carolina where the peak price exceeds the
trough by only 15%. The most volatile states are concentrated in New England and along
the North Eastern seaboard (Massachusetts, New York, New Jersey, Rhode Island,
Connecticut, New Hampshire, and Maine) and in California and Hawaii. While we will
use a continuous measure of volatility in our analyses below, for descriptive purposes we
label these the volatile states.
To exploit regional and time series differences in volatility in house prices we
construct indices of volatility by computing the standard deviation of the change in the
log real house price index over the previous five years for each of the 50 US states and 12
UK regions for which we have house price indices. These indices, which measure percent
volatility over the sample are plotted in Figures 4a and 4b, grouped by the same two
‘volatile’ and ‘non-volatile’ areas as before. Two things are important to note. First, the
higher levels of volatility in the UK (even in the ‘non-volatile’ regions) are apparent.
Second, in both countries, it will be the state/regional level volatility index, not an
average across groups of regions that enters our empirical specifications.
III. Housing Choices in the Presence of House Price Risk and the Housing Ladder
In order to think about how the housing ladder might affect housing demand in
the presence of house price risk we use the concept of a minimum housing ‘need’ that
changes with family size. This need can then be thought of as increasing over the life-
cycle as individuals form into couples, have children and reach their maximum family
size. Central to our empirical modeling is the idea that these increasing housing needs
over the life-cycle interact with future house price risks to generate an insurance role for
housing consumption early in life.5 In this section we discuss the intuition behind this
idea, before moving on to testing the predictions of such a framework empirically.
In a standard model without house price risk housing demand would increase with
wealth but would also adjust to reflect the minimum necessary level of consumption. In
such a framework one could write housing demands in each period as a function of
adjusted lifetime wealth (i.e. the present discounted value of lifetime wealth net of the
discounted sum of minimum necessary levels of housing over the lifetime6), the real user
cost of housing services, and the minimum level of housing needs in that period. Any
future change in household demographic composition would simply act through its effect
on adjusted wealth. While the consumption of housing services may involve the purchase
of a house and an asset accumulation decision, the assumption of perfect credit markets
and certainty would yield this aspect of housing consumption unimportant in such a
setting. We need to generalize this model in order to incorporate house price risk and
consider the additional role of housing as a durable asset.
For ease of exposition we will assume the life-cycle profile can be represented by
the following sequence of three discrete life-stages: At stage D = 1 the individual is
living with his or her parents, at stage D = 2 he or she partners to form an independent
family unit, and at stage D = 3 the couple has had children and completed its family size.
This is a simplified demographic profile but represents effectively the upside of the
housing ‘ladder’ that we wish to capture in our model.7 For further simplicity we will
assume that the leaving home decision D=1 → D=2 simply concerns a decision over
whether to rent or own in the light of the possible increase in family size associated with
the arrival of children between D=2 and D=3.
Without price uncertainty the rent/own decision will be driven by transaction
costs of ownership as well as the desire for mobility, the potential tax advantage of a
mortgage, and any down payment rules or constraints on the multiples of income that
may be borrowed. For a household that expects to remain in their house for a reasonable
length of time, for example at D=3 (the top stage of the demographic ladder) owning is
the most efficient way of achieving a desired level (and type) of housing service – with
idiosyncratic tastes a renter can never commit to stay long enough to make it in the
landlord’s interest invest in the renter’s idiosyncratic tastes. Hence we will assume for
simplicity that all households will be owner-occupiers at D=3 and that this is known to
them at D=2.8
Before turning to the introduction of house price risk, there are two aspects of the
supply of housing services, which are relevant to our discussion. First, a more inelastic
supply will induce a larger sensitivity of house prices to changes in demand and, in
particular, to fluctuations in incomes of young first-time buyers. The second aspect
relates to the rental market —imperfections and/or regulation of the private rental market
may make it difficult for the young to use rental housing as the step between leaving the
parental home and acquiring a house.
The introduction of house price uncertainty into the model adds an important
distinction between ownership and renting which will enhance the desire to accumulate
housing wealth and thus the need to become an owner earlier in the life cycle—house
price risk generates an incentive to accumulate housing equity at D=2 before the family
is complete. At first sight this may seem a puzzle since accumulation of a risky asset
might normally be expected to decrease with the level of price volatility for a household
with risk-averse preferences. That usual result does not hold because of the vital
insurance role played by housing in early life in our framework. We argue this intuitively
below, but to back up this intuition, in Appendix II we simulate the predictions of a
simple three-period model with constant relative risk aversion preferences that allows us
to demonstrate more formally the effects on housing consumption profiles of changing
volatility, the changing steepness of the housing ladder and changing degrees of risk
At D = 2 there are two choices: how much housing to consume, and whether to
own or to rent. If house prices are variable and uncertain then, given the expected
increase demand as the household moves up the demographic ladder from D=2 to D=3,
housing equity will be an important source of insurance against future house price risk.
Indeed, in the absence of a financial instrument that could insure this house price risk
(which may well be defined at a very local level), holding housing early in life may be
the only insurance mechanism. The larger the uncertainty in house prices and the steeper
the increase in minimum housing needs over the life cycle, the more important is the
insurance aspect of housing equity.
Thus the key mechanism for these effects is the insurance role of housing in
period 2. If prices turn out to fall or stay the same then ownership will not, ex post,
dominate renting. Indeed if house prices fall there will be some loss to ownership.
However, because of the strongly declining marginal utility of consumption associated
with housing consumption in period 3 approaching the minimum necessary requirement,
insuring the risk of house price rises is more important than avoiding the risk of a house
price fall. To achieve this, the consumer needs to hold an asset whose return is correlated
with (local) housing prices. If such an asset is not available on the financial market the
insurance can only be achieved by purchasing the asset itself. Consequently, other things
equal, the higher the level of house price uncertainty the higher the incentive to become
an owner-occupier. In this context increasing minimum housing requirements or
increases in risk aversion are acting in a similar way to an increase in volatility. By a
straightforward extension of these arguments, individuals will also stay away from
endowment mortgages and refinancing of housing equity for non-housing consumption or
In summary, the decision to accumulate housing equity early in the life cycle will
be an increasing function of house price volatility for risk-averse households who expect
an increase in family size. In the absence of an equity market in local housing assets, this
demand for housing equity also enhances the decision to own.10
One further extension that needs to be discussed, since we endeavor to control for
it in the empirical analysis that follows, is geographic mobility. If individuals anticipate
residing in less volatile areas in period 3 then their demand for insurance is reduced (and
the insurance value of their housing equity in period 2 will be reduced also to the extent
that house prices are not perfectly correlated across regions). It is expected volatility at
D=3 (from the point of view of D=2) that drives the insurance motive. In the case of
individuals in D=2 anticipating moving to a ‘safe’ area at D=3, both these factors are
likely to play a reduced role, although they could still be important to some extent.
IV. The Empirical Relationship between Housing Choices and Risk
On the basis of our discussions in the previous section, and the numerical model solutions
presented in Appendix II, there are three principal predictions that we will test
empirically in this paper: (1) other things being equal, individuals should buy homes
earlier in more volatile areas; (2) young homeowners are less likely to consume capital
gains on housing through refinancing in more volatile areas; and (3) young homeowners
will consume ‘more’ housing in more volatile areas than their counterparts in less volatile
areas. In the following subsections we deal with each of the above predictions in turn.
IV.A. Age of Home Ownership
In the presence of a housing ladder, individuals living in places with more volatile
housing prices need to self-insure by buying their first home at a younger age. In the final
column of Table 3, we list for both the UK and US the proportion of individuals who are
homeowners, by age for a typical year—1994. These patterns do not depend critically on
the year chosen. The data are also presented separately for the volatile and non-volatile
areas in both countries. While average rates of home ownership are similar, there are
striking differences by age between the two countries. Home ownership rates amongst
young households are far higher in the UK than in the US, with differences of 10
percentage points for householders between ages 20-29 and 13 percentage points those
between ages 30-39. However, through middle age, homeownership rates converge so
quickly that US rates actually exceed those in the UK among older households.
Since prices are far more variable in the UK, these cross-country differences in
home ownership rates are consistent with our theoretical implication that ownership
should occur at a younger age in more price volatile housing markets. However, when we
compare home ownership rates between the volatile and non-volatile areas within each
country, the challenge to our theory becomes more apparent. In both countries, owning a
home is somewhat less common among younger households in the volatile market.
However, there are other significant differences between these two markets in
each country that will presumably strongly affect the decision to own. Tables 4a and 4b
lists some of the more salient ones. Perhaps, most important, housing prices are much
higher in the volatile markets. For example, the average price of a home in the more
volatile states is almost twice that in the less volatile ones, which should certainly
discourage home ownership among the young. While rental prices are also higher in the
more volatile states, the percentage difference is 46% compared to 68% for housing
prices. Young individuals living in the volatile states also have more education,
household income, and are less likely to be married and to have children. All of these
factors are obviously relevant to the housing tenure decision so the final verdict on the
theory requires multivariate modeling.
In our multivariate analysis, we estimate a probit model of whether or not one is a
homeowner using a sample of individuals who are between the ages of 21 and 35. Results
are similar if one uses a somewhat younger or somewhat older age band that corresponds
to the rising part of the housing ladder. In addition to our measure of housing price
volatility described above, this model includes several relevant demographic attributes—
a quadratic in age, indicator variables for whether one is married and whether one has
children, the log income of the tax unit in which the individual participates, and measures
capturing years of schooling. We measure area and age specific housing prices by using
the PSID and BHPS to compute mean housing prices and mean rents in each state/region
for owners and renters respectively, within broad age groups. These prices as well as
benefit unit income are entered in logs.
The critical variable for testing our theory concerns housing price variability,
which varies across space and time. We construct a five-year moving window of the
standard deviation of the year-to-year differences in the log real housing prices in a
region11 as described in the previous section. Since our US housing price series starts in
1974, this means that our PSID analysis starts with the 1980 PSID and extends to the
1997 PSID. Since fewer historical years are available in the BHPS, the analysis there
covers the years 1991-2003.
As noted earlier, expected capital gains are likely to be an important component
of the demand for a risky asset like housing. Expected capital gains reduce the user cost,
reflecting the risk-return trade-off. To construct an expected gains variable we use the
change in the regionally varying log real house price index over the previous five years.
Precisely the same five-year moving window for house prices we use in constructing the
house price risk variable.
To control for the possibility that the variability in housing prices across regions
and states may simply be capturing unmeasured differences across states and regions, we
estimated all models with and without state and region effects. A linear time trend is
added to our models so our time series variation is relative to a common linear trend.
The results are displayed in Tables 5a and 5b, which lists marginal effects and
standard errors of all variables obtained from probit models. In both countries, we find
positive income effects (slightly higher in the UK) and education effects (a possible
proxy for permanent income) on home-ownership. Not surprisingly, marriage in both
countries encourages home ownership and at least in the US children do likewise. In the
US and the UK, we also have statistically significant negative price level effects on the
probability of owning a home. We also find a positive impact of expected capital gains,
although this is not uniformly significant across all model specifications.
In both countries, high area-specific rents also discourage home ownership. While
this may at first blush seem counter-intuitive, it is important to remember that there are
three options open to young persons in terms of their housing choices—owner, renter, or
living with others—especially parents. When we estimated models for whether one was a
renter, higher rental prices discouraged both renting and home owning.
The coefficients on the price volatility variables form the basis of the test of our
central prediction. In both the US and UK, we estimate statistically significant positive
effects of price volatility indicating that as predicted individuals choose to own homes at
a younger age in the more housing price volatile areas. When state/region dummy
variables are included, these estimated effects are remarkably similar in the two countries
so that on the margin Britons appear to react more only because volatility on average is
so much higher there.
IV.B. The Decision to Refinance
As discussed above, our key hypothesis is that households in areas where housing prices
are volatile should self-insure at young ages by holding more housing. However, if they
were to buy a house and then refinance and use the proceeds to finance consumption or to
purchase risky assets this would simply undo the safety housing provides. As such, we
would expect less of such behavior in volatile areas and we test this prediction in this
section. Although imperfect, our two datasets provide some measure of the extent to
which individuals engage in such activities. With regard to the US, PSID data contain no
direct questions in each year on refinancing, so we define an indicator of refinancing to
take the value 1 if an individual’s mortgage is observed to have risen by a specified
amount between waves.12 The problem with this measure is that individuals could well be
using the extra finance to improve their home, which would not unravel the housing as
price insurance mechanism, thus making it an imperfect measure for our purposes.
This prediction can, however, be directly addressed in the UK using BHPS data,
where individuals are asked specific questions about whether they refinanced their
housing equity between waves, and if so whether the purposes for which the resulting
money was used. With such detailed questions we are able to construct a more precise
indicator in the UK that takes the value 1 only if individuals refinance between waves and
do not increase the quantity or quality of housing as a result.
Our results are summarized in Tables 6a and 6b. In addition to the non-price
variables that were part of the home ownership model, we included a measure of home
equity in the previous year to capture the amount available for refinancing. In both
countries, using both measures of refinancing, the predictions of the theory are borne
out—individuals in more risky areas are less likely to refinance, conditional on other
characteristics and their initial level of net housing equity.
IV.C. Increased Consumption of Housing
As pointed out in Section 2, one can insure against future housing price volatility in
period D=3 not only by purchasing a house in period D=2 but also by consuming more
owned housing than one might otherwise want given the objective demographic
circumstances. Moreover, in the presence of borrowing constraints there is a possibility
that, if prices rise more quickly than income, debt to income restrictions may prevent
individuals being able to purchase a larger home at D=3. With this possibility on the
horizon individuals, already more likely to be an owner-occupier as a result of the
increased volatility, would also choose to increase their consumption of housing, since in
the case of prices rising the capital gain will be higher and can be used as down payment
on the final home in order to offset the debt to income restriction. Indeed, in the UK, the
two conditions are often linked (since on a secured loan the consequences of default to
the lender are reduced with a higher down payment) such that individuals with higher
down payments can borrow a higher multiple of income.
In order to measure the consumption of housing, for the purposes of testing this
prediction we use two variables—the number of rooms in the house, and the gross value
of the house.13 Neither is perfect since the former omits possible quality effects and the
latter may be contaminated by unmeasured price variation leading to uncontrolled for
demand effects. Nevertheless, each provides a useful complementary test for the
predictions of the model. For each of these measures of housing consumption, we use a
standard Heckman type selectivity model to evaluate the predictions for home-owners
only, using the probits reported in Tables 5a and 5b as the selection equations and
omitting the rental price from the continuous part of the model.
Tables 7a and 7b report the results of estimating selection models for the number
of rooms occupied by young homeowners. These estimates show significant positive
effects of volatility on house size but only in the UK—young British home owners in
risky areas tend to consume more rooms than their counterparts in safer areas in order to
partially insure themselves against housing price risk. The effects are positive in the US
as well but not statistically significant at conventional test levels.
Other estimated parameters accord with a priori intuition. The number of rooms
increases with income, education, whether an individual is married, and with the presence
of children, and decreases with the average price of housing per room in the area. The
magnitude of the demographic effects (marriage and children) and the income effects are
similar in the two countries. Finally, those individuals moving from risky to safe areas
have a reduced number of rooms, as would be predicted by their insurance motive being
reduced, although not by enough to offset the volatility effect altogether.
In Tables 8a and 8b we repeat this analysis using gross house value as our
measure of housing consumption. Again in both countries, as predicted by our theory
individuals in risky areas choose to have higher housing wealth than those living in safe
areas. This effect is reduced for those observed to move from risky to safe areas during
the period of our data. Thus, those individuals who end up moving out of the risky
housing price areas appear to insure less in the sense that they do not over consume
housing when they are young. Once again, the principal demographic variables enter with
the expected signs and in about the same magnitude in both countries—home values
increase with marriage, children, and age (at least until middle age). Similarly, income
and education effects are positive in both countries although our estimated current
income elastic city is much higher in the US than in the UK.
The models estimated in Tables 7 and 8 are based on two alternative and
imperfect measures of housing consumption. However, the general similarity of the
estimated models across both specifications, and in particular the similar estimated
effects of our measure housing price variability on housing consumption in both countries
lends support to the predictions of our model.
IV.D Endowment Mortgages
Over the period covered by our data, one relatively common financial instrument used to
finance house purchases in Britain was an endowment mortgage. During the life of the
mortgage, the borrower makes only interest payments on the loan, leaving the principal to
be repaid at the end of the term of the mortgage. In addition to the interest, the borrower
pays into a saving scheme, which is designed to mature and repay at least the value of the
capital sum borrowed at the end of the period of the loan. Throughout the 1980s and
1990s these schemes were common, with the most common type of saving scheme being
an endowment policy that is an investment product—essentially term life insurance with
the fund invested in the stock market.
Because of the risky nature of this product, our framework would predict that
households who live in volatile areas should be less likely to choose this type of
mortgage.14 These predictions are borne out using the same framework as the tests
presented above. In Table 9, we report results obtained from probit models with the
dependent variable being whether individuals finance their house purchase with an
endowment mortgage. Since mortgage arrangements typically do not change over the
term of the mortgage (and in the case of endowment policies the penalties for early
termination are high), we are able to use homeowners of all ages for this test, thus also
implicitly increasing the period over which are effects are apparent. Whether or not we
include region dummies, British families who live in more volatile housing price areas
are less likely to take out an endowment mortgage. This estimated effect is statistically
Typically, risk-averse individuals will avoid risky assets as volatility increases. In this
paper we show that owner-occupied housing is an exception to this rule. The
consumption role of housing wealth, coupled with increasing necessary levels of housing
over the life cycle due to demographic changes, and the fact that individuals will typically
prefer to own rather than rent, mean that individuals will expect to be consuming a risky
commodity—owner occupied housing—in middle age. Since housing is a necessity the
utility consequences of this risk might be expected to be relatively large. In the absence
of suitable financial products to insure this risk, this will lead individuals to invest in
housing early in the life cycle as a way of insuring future price fluctuations. Not only
does this lead to higher owner-occupation rates, it also leads to more housing wealth and
less propensity to realize capital gains on housing through refinancing to fund non-
Using micro data from two countries we have constructed tests of these
predictions and all are borne out empirically. Cross-country differences between the US
and UK correspond to the cross-country differences in volatility—the UK is more volatile
and UK households own earlier, and have more of their portfolio in housing. Because this
may be driven by other differences between countries, we use within country tests that
rely on time-series and cross-sectional variation in volatility within and across states (in
the US) or regions (in the UK) we continue to find empirical support for the predictions
of the theory.
Banks, James, Richard Blundell, and James P. Smith. 2003. Wealth portfolios in the US
and the UK. In D.Wise (ed.), Perspectives on the Economics of Ageing, Chicago
University Press, pp 205-246.
Bhatia, Kul. 1987. Real estate assets and consumer spending. Quarterly Journal of
Browning, Martin, and Annamaria Lusardi. 1996. Household saving: Micro theories and
micro-facts. Journal of Economic Literature 34:1797-1855.
Brueckner, Jan. 1997. Consumption and investment motives and the portfolio choices of
homeowners. Journal of Real Estate Finance and Economics 15:159-180.
Calhoun, Charles A. 1996. OFHEO House Price Indexes: HPI Technical Description.
Web address: http://www.ofheo.gov/media/archive/house/hpi_tech.pdf
Campbell, John, and Joao Cocco. 2005. How do house prices affect consumption?
Evidence from micro data. NBER Working Paper 11534; forthcoming in Journal of
Cocco, Joao. 2005. Portfolio choice in the presence of housing. Review of Financial
Davidoff, Thomas. 2006. Labor incomes, housing prices, and home ownership. Journal
of Urban Economics 59:209-235.
Engelhardt, Gary. 1996. House prices and home owner saving behavior. Regional Science
and Urban Economics 26:313-336.
Engelhardt, Gary. 1996. Consumption, Down payments, and liquidity constraints.
Journal of Money, Credit, and Banking 28(2):255-271.
Ermisch, John, and David J. Pevalin. 2004. Early childbearing and housing choices.
Journal of Housing Economics 13:170-194.
Flavin, Margorie, and Takashi Yamashita. 2002. Owner-occupied housing and the
composition of household portfolio over the life cycle. American Economic Review
Ioannides, Yannis, and Stuart Rosenthal. 1994. Estimating the consumption and
investment cemands for housing and their effect on housing tenure status. Review of
Economics and Statistics 76(1):127-141.
Ortalo-Magne, Francois, and Sven Rady. 2004. Bargaining over residential real estate:
evidence from England. Journal of Housing Economics 56:192-216.
Ortalo-Magne, Francois, and Sven Rady. 2006. Housing market dynamics: On the
contribution of income shocks and credit constraints. Review of Economic Studies
Poterba, James M. 1991. House price dynamics: The role of tax policy and demography.
Brookings Papers on Economic Activity 2:143-183.
Sinai, Todd, and Nicholas S. Souleles. 2005. Owner-occupied housing as a hedge against
rent risk. Quarterly Journal of Economics 120(2):763-789.
Sheiner, Louise, and David Weil. 1992. The housing wealth of the aged. National Bureau
of Economic Research Working Paper No. 4115.
Skinner, Jonathan. 1994. Housing, wealth and aggregate aaving. Regional Science and
Urban Economics 19:305-324.
Venti, Stephen F., and David A. Wise. 2001. Aging and housing equity: Another look. In
D. Wise (ed.), Labor incomes, housing prices, and home ownership. Chicago
University Press, 127-180.
1 A detailed data description is provided in Appendix I
2 In the UK there is little evidence of cohort effects during the early part of the adult life cycle for the
period 1968-1998 (Banks, Blundell, and Smith, 2003). This suggests the rise would be the same whether
we look at individual date of birth cohorts or pool across cohorts as in the tables here. In the US, there is
some evidence of the number of rooms plateauing out at higher values among more recent cohorts.
3 The profiles in Table 1 show some evidence of ‘downsizing’ at older ages as children move out and the
parents transit into retirement. While this downsizing may be important especially for retired American
households (see Venti and Wise, 2001), it is not the focus of this paper, which concentrates instead on the
implications of the steps up the ladder earlier in life and a full analysis would need to take into account the
possible effects of cohort differences amongst older those at older ages on these profiles.
4 Completed family size is computed by taking individuals aged 50 or over and assuming they will not have
any more children. We then look back through their fertility history and find the age at which their final
child was born, and call this age the age of completed family size.
5 In a related framework, Francois Ortalo-Magne and Sven Rady (2002) have looked at the theoretical
predictions of an equilibrium model of home-ownership when house prices are volatile.
6 This wealth variable contains the current value of assets and the future stream of discounted income
flows. Housing equity and other assets will be added in our discussion of uncertainty below.
7 We ignore here older stages of the life cycle where the possibility of downsizing comes into play (see
Stephen Venti and David Wise (2001) and Lousie Sheiner and David Weil (1992) for example).
8 To the extent that this probability is less than one then any insurance motive will be dampened but as long
as the positive probability of homeownership at D=3 is not zero the insurance motive will still exist. Since
our empirical tests are simply for the presence of an insurance effect of house price risk on housing choices
at D=2 all they formally require is that this probability is not zero.
9 Borrowing constraints add further refinements to the model. They typically take two forms: a down
payment constraint and a multiple income (or debt to income) constraint. The down payment is
proportional to the house price. The multiple income constraint restricts the mortgage to be a multiple of
current income. With such constraints in place, the potential downside of a house price rise between D=2
and D=3 for a non-owner enhances the insurance value of ownership at D=2. If house prices rise relative to
incomes then the capital gain reduces the mortgage requirement and makes it more likely that the earnings
to mortgage debt can be met. Such borrowing constraints add to the insurance value of ownership since an
unexpected price increase at D=3 considerably relieves the down payment constraint.
10 An additional reason for ownership is given by rental price risk. As Sinai and Souleles (2005) point out,
house ownership insures housing consumption from rental price risk (although it may not alleviate cyclical
fluctuations in housing costs when variable rate mortgages are the predominant form of finance for housing
purchases). Our focus here is specifically on the housing ladder where we show house price risk enhances
the probability of ownership and the speed with which an individual moves up the ladder. At this stage of
the life-cycle where expected duration of stay in rental housing is relatively short, rental price risk may be
less relevant than for lifetime renters. In addition, young agents can avoid rental price risk by living with
their parents until they are ready to buy a home. This is relatively common pathway in Britain.
11 For each of the 50 US states and the 12 UK regions.
12 In practice, small rises could simply be a result of measurement error, so we choose a variety of
thresholds above which we assert a change in mortgage can be interpreted as a refinance. The specification
in Table 6a uses a definition of mortgage rising by at least 30% or $5000, whichever is the greater.
13 With increasing availability of appropriate panel data on wealth, there has been renewed interest in the
study of housing wealth dynamics and its implications for other economic factors. Flavin and Yamshita
(2002) look at the effect on household’s optimal financial asset holding of integrating housing (i.e. both
housing wealth and the associated consumption demand for housing services) into the portfolio model. In a
more empirical study, Banks, Blundell and Smith (2003) show that housing wealth differentials between
the US and the UK offset to some extent the differences in financial wealth observed between the two
countries. But in spite of recognition of the dual importance of housing as both consumption and
investment, the implications of the often-considerable housing price uncertainty for the life-cycle path of
housing wealth are not well understood.
14 One complication in testing this prediction is that particularly in the 1980’s (and early 1990’s), there is
some evidence that mis-selling of this type of mortgage took place by mortgage providers. In particular,
there is the possibility that consumers were not fully informed of the nature of other choices of mortgage
arrangements available or about the risky nature of the endowment policy. Assuming such effects were
constant across regions, however, we might still expect those living in more volatile regions to be less
likely to take out such mortgages.
Appendix I: Data Sources
The PSID started in 1968 collecting information on a sample of roughly 5,000 (original)
families. Of these, about 3,000 were representative of the US population as a whole (the
core sample), and about 2,000 were low-income families (the Census Bureau's Survey of
Economic Opportunities sample). Thereafter, both the original families and their split-
offs (children of the original family forming a family of their own) have been followed
giving a total of around 35,000 individuals. Panel members were interviewed each year
until 1997 when a two-year periodicity rule was established. All original members of the
1968 households and their progeny are considered sample members and thus are part of
the panel even if they move out of the original household. The US models presented in
this paper include the SEO over sample although they were also estimated using only the
core sample and our results regarding the effects of housing price volatility were not
In each wave of the panel, the PSID asks detailed questions on individual and
household income, family size and composition, schooling, education, age, and marital
status. State of residence is available yearly and individuals are followed to new locations
if they move. Unlike many other prominent American wealth surveys, the PSID is
representative of the complete age distribution. Yearly housing tenure questions
determine whether individuals currently own, rent or live with others. Questions on
housing ownership, value, and mortgage were asked in each calendar year wave of the
PSID.1 Renters are asked the amount of rent they pay and both owners and renters are
asked the total number of rooms in the residence.
1 Mortgages are not available in the PSID for years 1973, 1974, 1975, and 1982.
In addition to the PSID, housing price data were obtained from the Office of
Federal Housing Enterprise Oversight (OFHEO) House Price Index. These data contain
quarterly and yearly price indexes for the value of single-family homes in the US in the
individual states and the District of Columbia.2 These data use repeat transactions for the
same houses to obtain a quality constant index and is available for all years starting in
1974. All yearly housing prices by state are reported relative to those that prevailed in
1980. By 1995 there were almost 7 million repeat transactions in the data so that the
number of observations for each state is reasonably large. No demographic data are
available with this index.
For the UK, we use the British Household Panel Survey (BHPS). The BHPS has
been running annually since 1991 and, like the PSID, is also representative of the
complete age distribution. The wave 1 sample consisted of some 5,500 households and
10,300 individuals, and continuing representativeness of the survey is maintained by
following panel members wherever they move in the UK and also by including in the
panel the new members of households formed by original panel members. The BHPS
contains annual information on individual and household income and employment as well
as a complete set of demographic variables. Like the PSID, data are collected annually on
primary housing wealth, and on secondary housing wealth.3
In addition to the BHPS, regional house price data were obtained from the
Nationwide Building Society House Price series, which is a quarterly regional house
price series going back to 1974. Rather than use a repeat sales index, the prices are
2 For details on this data see Charles Calhoun (1996). The paper is available on the OFHEO website.
3 Housing wealth and mortgages are not available in 1992.
adjusted for changes in the mix of sales to approximate a composition constant index, and
are also seasonally adjusted.
Throughout the paper we take care to define the unit of analysis as the benefit unit
(i.e., singles or couples with dependent children) such that young individuals at the
beginning of the life-cycle living in shared accommodation or with other family members
are not lost from the analysis as subsidiary adults in households headed by other
individuals. This is particularly important for older independent children who are still
residing with parents and who would show up in middle-aged households in a
conventional head of household based analysis. In both countries, housing wealth is
allocated to the home owning benefit unit only. Hence a 25-year-old living with their
parents in an owned property is not defined as an owner (unless they own the property
jointly with their parents) and is assigned zero housing wealth.
We use several housing wealth concepts in this paper. The current value of the
house is derived in both the PSID and BHPS by asking respondents to report the current
market value of their home while housing equity is constructed by subtracting from the
current house value the outstanding mortgage.
Appendix II: Numerical simulation of a simple model of ownership and housing equity in
the presence of house price risk and a housing ladder
The integration of housing price risk into a single theoretical framework is
complex and even algebraic closed form solutions will only be possible under certain
(restrictive) forms of preferences. Ideally, however, we want to use relatively flexible
preferences for consumption and housing to generate predictions relating to the effects of
house price risk. In this appendix we use numerical methods in order to offer insight into
the predictions of the model using a very simple set of specifications for preferences, the
steepness of the housing ladder, and the time-series process for the underlying
For the purpose of our simulations, we assume that individuals maximize
expected discounted life-time utility, with the utility functions for an individual in each of
the decision periods being given by:
tq is the consumption of housing services in period t and all other consumption is
tc . To accord with our discussions of section III, these preferences are
characterized by having a necessary level of housing consumption,
tq , in each period to
capture the housing ladder; but they also take the CRRA form to allow us to look at the
impact of varying risk aversion on the predictions of the model.
4 Ultimately, many other extensions could be looked at with this approach, such as the sensitivity of
predictions to rental premia, the cost of mortgage borrowing, the extension of the model to a greater
number of time periods or the differences in predictions that emerge as we allow income uncertainty (with
differing degrees of correlation between income and house price shocks). But we leave these extensions for
further work since, at this stage, we want to make the model as simple as possible whilst still remaining
sufficiently general to examine the specific predictions on which the empirical analyses in this paper are
We solve the numerical model with three periods, aimed at capturing the phases
of the life cycle discussed in Section III, rather than calendar years, quarters or even
months. When building a numerical solution algorithm, the choice of units and parameter
values forces one to think carefully about the relative length of periods. In taking
numerical methods to our model we essentially need to think of periods of unequal
length, in order to capture the sense in which period 2 (the middle rung on the housing
ladder) is a transition to a more permanent state of completed family size and a
‘permanent’ family home. A convenient way in which to do this is to introduce factors 2
and 3 , with 0 < t ≤ 1, t =2,3 and 2 ≤ 3 , which describe the flow of consumption
services qt from housing stock Ht , so that qt = tHt.
We choose a stylised model in which the only uncertainty is in house prices. In
accordance with our earlier discussions, we assume that in period 1 everyone is a renter
and in period 3 everyone is an owner. The key decision is whether to own in period 2 or
wait until period 3. We show that increasing house price uncertainty increases the pay-off
to ownership in the second period. This pay-off is larger the larger the degree of risk
aversion and the stronger the gradient in the housing ladder. As we are only interested in
the relative pay-off of ownership we normalise on first period utility and examine relative
pay-off in periods 2 and 3. The budget constraint for periods 2 and 3 under each option is
(2a) [Owner at t=2]:
2 3 3
yyp p HccHH
(2b) [Renter at t=2]:
2323222 3 3
depending on which tenure is chosen, where yt are discounted incomes, pt are discounted
prices, ct are discounted consumptions, and is the rental premium.
Implicit in this set up is that an individual can borrow or save at the same (safe)
rate of interest equal to the discount rate. Finally, we introduce house price uncertainty in
period 3 by allowing p3 to take the value p2(1+) with probability ½ and p2(1-) with
probability ½. We can then vary the variance of housing price uncertainty by solving the
model for different values of .5
We solve the model by backward induction with a relatively straightforward
numerical method that involves a discrete grid search across all possible paths for
housing consumption in each period, q, consumption in each period, c, and the
owner/renter decision in period 2. For the purposes of the solution, baseline values are set
at:, = 1, = 0.3, 2 = .5, 3 = 1,
0q y3 = 200 and y 2 = .5y3. The later equality equates
the flow of income across the two periods given the choice of 2 and 3. The model is
then solved under varying degrees of uncertainty for various values of the necessary level
of housing in period 3 (which we shall refer to as D) ranging from D=10 to D=40, and for
various values of the risk aversion parameter, .
Figure A1.a shows the difference between the expected utility of renting and
owning in period 2, expressed as a fraction of the utility of renting, as the variance of
housing prices increases and as the minimum level of housing required in period 3, i.e.
the steepness of the housing ladder, increases. The figure shows that increases in the
minimum level of housing demand in period 3 result in an increase in the relative utility
of owning in period 2 for all positive levels of volatility. Similarly, for all levels of the
minimum housing requirement in period 3, increasing price volatility results in a stronger
preference for owning: Increasing house price risk reduces expected utility for both
5 In this discussion we abstract from expected capital gains. However, our empirical model will allow for a
capital gains term which will reflect the risk-return trade off. Holding the riskless return constant, an
expected capital gain will reduce the user cost of housing and make ownership more attractive.
renters or owners in period 2 but the impact is stronger on the rental option. Consequently
there is a gain in expected utility terms from ownership in period 2 and this gain increases
with risk. Figure A1.b presents a complementary analysis but where we hold the housing
ladder constant and vary the degree of risk aversion in preferences. As risk aversion
increases the slopes of the profiles with respect to volatility steepen.
In addition to the home ownership predictions the model should also have
predictions for the quantity of housing consumed as discussed in section III. Figures A2.a
and A2.b show the predictions for housing consumption in period 2 as the housing ladder
steepens and as risk aversion increases. Figure A2.a shows that, for any level of the
minimum housing requirement in period 3, as volatility increases the quantity of housing
demanded in period 2 increases—individuals buy more insurance as risk accumulates.6 If
volatility is significant, a steeper housing ladder results in more housing consumption in
period 2. This implies that not only will individuals be more likely to purchase a house in
period 2, they will also be more likely to purchase a ‘bigger’ house. Note that for the very
lowest value of the minimum housing requirement (D=10) the quantity of housing
actually declines with volatility. At such a low value of the minimum (and given the
relative preference for housing implied by our choice of α of 0.3) the housing ladder
constraint is not effectively binding and therefore the predictions of the model are in
accordance with the standard case: individuals choose less of a risky activity.
Figure A2.b presents similar results by risk aversion coefficient. Once again, as
risk aversion increases, the quantity demanded of housing in the second period increases.
6 Varying the minimum housing requirement and keeping life-time resources constant also generates a
wealth effect. This is not important for our empirical tests since we will be examining demand for housing
as volatility varies for a given steepness of the housing ladder. As a result we abstract from this wealth
effect in this figure by normalizing the housing demand to its zero-volatility value in the two figures.
While not shown in these graphs, our model also has implications for non-housing
consumption in period 2, which is generally declining in housing price volatility.
Table 1. The Number of Rooms by age of head of household
Age of head of household
Owners and Renters
Owners and Renters
3.694.45 4.984.89 4.07
3.924.69 5.24 5.17 4.54
Pooled data from the PSID and BHPS. US data excludes bathrooms, UK data excludes kitchens
Table 2. Changes in rooms for movers, by type of buyer
Age of head of benefit unit
First Time Buyers - Before
First Time Buyers - After
First Time - Difference
Repeat Buyers - Before
Repeat Buyers - After
Repeat - Difference
First Time Buyers - Before
First Time Buyers - After
First Time - Difference
Repeat Buyers - Before
Repeat Buyers - After
Repeat – Difference
Note: pooled PSID and BHPS data from 1990-1999 and 1991-2003 respectively. First time
buyers restricted to those previously living in rented accommodation. Cell sizes too small in UK
for age < 25.
Table 3. Proportion of individuals who are homeowners in 1994
Age ‘Volatile’Non-volatile All
0.652 0.641 0.648
0.583 0.649 0.633
Data are from the 1994 BHPS and PSID.
Table 4a. Differences across broad regions, 21-35 year olds, US
Fraction of population (1999)
Ever had a child
Years of education
Log income in 1995$
Mean PSID house value
Mean PSID annual rent
Table 4b. Differences across broad regions, 21-35 year olds, UK
Fraction of population (2000)
Has a child
Education – low
Education – medium
Education – high
Ln income (in £ 2000)
Mean BHPS house value (£)
Mean BHPS weekly rent (£)
Source- PSID and BHPS.
Table 5 a. Probability of Home-Ownership, US
Ever have a child
Ln Housing Prices -0.0559
Exp. Capital Gains 0.0360
Ln Rental prices
Ages 21-35. Models also control for city size, missing values, trend and
number of waves.
Std. ErrdF/ dx
Table 5b. Probability of Home-ownership, UK
Educ - low
Educ - medium
Ln House Prices
Exp. Capital Gains
Ln Rental Prices
Std. ErrdF/ dx Std. Err
Ages 21-35. Models include controls for living in a big city, number of waves observed in panel,
No No Yes
Table 6a. Probability of Refinancing a US Home
Price Volatility Index -0. 5043
Ever have a child
Ln House Equityt-1
Ages 21-35. Models also city size controls, missing value dummies.
Table 6b. Probability of refinancing a UK home
Educ - low
Educ - medium
Ln equity t-1
Ages 21-35. Models include controls for living in a big city, number of waves
observed in panel, trend, tax unit composition change between waves t-1 and t.
Table 7a. Number of rooms in US
coeffStd. Errcoeff Std. Err
Price Volatility Index
Ever have a child
Ln Housing Prices
Ages 21-35. Models also city size controls, trend, missing value dummies,
number of waves observed in panel. Selection equation is reported in Table 4.1a.
rental price omitted from rooms equation.
Table 7b. Number of rooms in the UK
Educ - low
Educ - medium
Ln House Price
Ages 21-35. Model also includes controls for city, trend, number of waves
observed in panel. Selection equation is reported in Table 4.1a. rental price omitted
from rooms equation.
coeff Std. Err
Table 8a. Gross housing wealth in the US
Price Volatility Index
Ever have a child
Ln Housing Prices
Ages 21-35. Models also city size controls, missing value dummies, number of
waves observed in panel. Selection equation is reported in Table 4.1a. rental price
omitted from rooms equation.
Table 8b. Gross housing wealth in the UK
CoeffStd. Err Coeff Std. Err
CoeffStd. Err CoeffStd. Err
Educ - low
Educ – medium
Ln House Prices
Ages 21-35 only. Models also includes controls for city, trend, number of waves
observed in panel. Selection equation is reported in Table 4.1a. rental price
omitted from rooms equation.
Table 9. Probability of holding endowment mortgage, home owners in UK only
dF/ dxStd. Err
Education - low
Education - medium
All ages. Models also include number of waves observed in panel, city trend.
Figure 1a: The demographic ladder, US
Figure 1b: The demographic ladder, UK
2025 3035 4045 5055 60
Ever had child over age 5
Completed family size
Own child over 5 in household
Figure 2. Comparison of UK and US house prices
Figure 3a: US mean house price index by area, 1980-1997
Figure 3b: UK mean house price index by area, 1980-2000
Figure 4a. Regional volatility indices, US, 1980-1997, by area
Figure 4b. Regional volatility indices, UK, 1980-2000, by area
198019811982 19831984 19851986 1987 1988 1989 1990 19911992 19931994 19951996 1997
49 Download full-text
Figure A1a: Relative utility of owner occupation when young
by variance of house prices and steepness of housing ladder
0 0.050.1 0.150.2 0.25
House price variance
Relative utility of owning
Figure A1b: Relative utility of owner occupation when young
by variance of house prices and degree of risk aversion
0 0.050.10.15 0.2 0.25
House price variance
Relative utility of owning
Gamma=1 Gamma = 2 Gamma = 4Gamma = 8