Price dynamics in public and private housing markets in Singapore
In down-payment constrained housing consumption models, increases in house prices could trigger household mobility decisions in housing markets. This study empirically tests house price dynamics associated with the mobility of households in the public resale and private housing markets in Singapore.
Published on: Mar 4, 2016
Transcripts - Price dynamics in public and private housing markets in Singapore
Price Dynamics in Public and Private Housing Markets in Singapore
Department of Real Estate
National University of Singapore, Singapore
Department of Finance
Southern Taiwan University of Technology, Taiwan
Department of Finance
National Sun Yat-sen University, Taiwan
Date: September 16, 2004
Revised: September 19, 2006
In down-payment constrained housing consumption models, increases in house prices
could trigger household mobility decisions in housing markets. This study attempts to
empirically test house price dynamics associated with the mobility of households in
the public resale and private housing markets in Singapore. The results show that
stochastic permanent breaks were found in the public housing resale prices and
private housing prices. The relative prices drift apart occasionally, but mean-revert to
a long-run fundamental equilibrium. Error correction mechanisms and lagged public
housing prices were also found to have significant explanatory effects for price
changes in the private housing markets. The results support the hypothesis that
household mobility creates co-movements of prices in public and private housing
submarkets in the long run.
Keywords: Housing price dynamics, household mobility, stochastic permanent
* Corresponding author. Address: Department of Real Estate, National University of Singapore, 4
Architecture Drive, Singapore 117566. Email: firstname.lastname@example.org. We wish to thank the editor,
Henry Pollakowski, Sau Kim Lum, Seow Eng Ong, Yuming Fu and an anonymous referees for
their constructive comments and suggestions.
Price Dynamics in Public and Private Housing Markets in Singapore
Housing is both an investment as well as a consumption good. Service flows derived
from housing are durable. Households do not trade housing goods as frequently as
other non-durable and financial goods because of the high initial equity outlay
involved. Liquidity constrained households, however, trade-off consumption of other
non-durables for housing goods. Increases in housing prices generate positive wealth
to households, which is translated into positive consumption of housing and non-
housing goods by households (Case, Quigley and Shiller, 2005). 1
The households’ mobility decision in trading up starter flats for bigger and higher
quality houses is dependent on the housing wealth accumulated (Stein, 1995; Ortalo-
Magne and Rady, 2004, 2006). House price changes create self-reinforcing effects
that run from down payments, to demand and then back to housing prices (Stein,
1995). The chain reaction also occurs vertically across different housing segments on
a property ladder (Ortalo-Magne and Rady, 2004, 2006). Credit constrained
households will move, if price increases in their existing houses were large enough to
cover outstanding mortgage balances and down payments for new houses. In a
declining market, constrained households are more reluctant to move. They will
instead “fish” for above-market sale prices.2 As a result, transaction volume will drop,
and housing prices will be adjusted downward.
In Singapore, the housing market is distinctively divided into two sub-segments. The
public housing market constitutes about 80% of the total housing stock (Phang, 2004),
and houses more than 90% of the population in Singapore. The supply of new public
housing units is regulated by the government through the Housing Development
Board (HDB). Public housing is an important source of housing wealth for liquidity
constrained households in Singapore. Private housing sales in Singapore are closely
related to increases in public housing wealth (Bardhan, Datta, Edelstein and Lum,
In Singapore, private housing wealth is transitory. Phang (2004) found no significant positive
effects of private housing wealth on aggregate consumption. However, Edelstein and Lum (2004)
found that public housing wealth was significant in affecting households’ consumption.
The “fishing” behavior of constrained households is rational, if and only if the price process is
expected to mean revert in the long run (Case and Shiller, 1989, 1990).
2003). Public housing policies were also found to have significant impact on private
housing price dynamics (Phang and Wong, 1997; Lum, 2002; Ong and Sing, 2002).
Do the property ladder and housing wealth models (Ortalo-Magne and Rady, 2004,
2006) explain the mobility behavior of Singaporean households? The hierarchical
segmentation of the housing market in Singapore offers a unique housing ladder
structure for the test of household mobility decisions. In Singapore, the public housing
market is a highly regulated market, where new public (HDB) houses are sold at
concessionary prices to eligible households. There is also an authorized secondary
public housing market established since 1971 (Phang and Wong, 1971) through which
households could trade their public houses after fulfilling a 5-year minimum
occupancy requirement. Price differences that exist between the new and secondary
HDB housing markets create a permanent source of wealth to public house owners
(Bardhan, Datta, Edelstein and Lum, 2003). Capital gains in the public housing
market are an important source of down-payments that are used by households in the
purchase of private houses. On the other hand, trade-down options are not usually
exercised by households that own private houses in the upper rungs of the housing
ladder. This could be due to a strong bequest culture that is prevalent among
Upward mobility is more evidenced than the downward path on the housing ladder
by Singapore households. Therefore, co-movements in housing prices, if triggered by
housing mobility activities, are likely to be asymmetric across the housing ladder in
Singapore. The paper aims to empirically test the long-run co-movements between
public and private housing sub-markets in Singapore. The results offer useful a
empirical verification to theoretical propositions developed by studies on the property
ladder, housing wealth creation and mobility. This paper is organized into six
sections. Section 1 discusses the motivations of the study. Section 2 reviews the
housing literature relating to liquidity constraints and upward mobility. Section 3
describes the ladder structure of the Singapore housing market, and discusses possible
testable hypotheses. Section 4 discusses the empirical methodology and data. The
empirical results are presented in Section 5. Section 6 concludes with possible policy
2. Literature Review
Stein (1995) develops a theoretical housing life-cycle model to explain down-payment
effects on households’ mobility decision. Stein’s (1995) models show that house price
shocks create self-reinforcing effects that run from down payment to housing demand
and then back to house prices. Ortalo-Magne and Rady (2004, 2006) 3 show that
household mobility follows a well defined path from a starter flat to a trade-up house.
The households’ decision of moving up and down the property ladder is dependent on
their endowment and housing preference profiles. The ability of households to afford
down payments is a key driver of the price reaction to income shocks. When the down
payment constraint is binding on housing purchase, households’ consumption
behavior in the early phase of the cycle could be distorted (Slemrod, 1982). In the
respective models of Stein (1995) and Ortalo-Magne and Rady (2004, 2006), a
positive house price and transaction volume relationship is predicted in equilibrium.
Genesove and Mayer (1997) empirically tested the positive loan-to-value (LTV)
effects on prices and time to sale of houses. They showed that households with higher
LTV constraints sold their houses at higher prices, but over longer periods of time.
The “fishing” behavior of constrained households in the Stein model that causes a
decline in housing transactions in a market with falling house prices was also not
rejected in the Genesove and Mayer (1997) tests. Bardhan, Datta Edelstein and Lum
(2003) showed that the positive gains in public (starter) houses are a significant factor
that drives private housing market transactions in Singapore. Lee and Ong (2005) also
found consistent empirical evidence of the positive effects of housing wealth and LTV
constraints on the upward mobility decisions of a sample of households who own
public housing in Singapore.
The mobility of constrained households causes serial dependence in housing prices
(Stein, 1995; Ortalo-Magne and Rady, 2004 and 2006). Case and Shiller (1989, 1990)
found evidence to support the predictability of housing prices using historical price
changes and other fundamental variables. Price discovery between the public and
In Stein (1995), all the households are assumed to have owned a house at the initial time period,
and the households could only choose to move or stay put in the existing houses. The models of
Ortalo-Magne and Rady (2004, 2006) are more dynamic, which describe the inter-temporal
mobility of households from starter flats to trade-up houses over time.
private housing prices in Singapore was found by Ong and Sing (2002). They further
showed that the causality effects are bi-directional in the two housing markets.
The public housing built by the government’s agency, the Housing Development
Board (HDB), is the predominant housing type in Singapore. It constitutes about 88%
of the housing stock based on the census statistics in 2000. Government policies
affecting public housing, in particular those relating to the Central Provident Fund
(CPF) 4 , were found to have significant impact on prices of the relatively small private
trade-up housing market (Phang and Wong, 1997). The government’s role in the
supply of land for private houses also created significant dampening effects on private
housing prices (Lum, 2002). Housing wealth created in the public housing market was
more sustainable compared to private housing wealth. Gains in public housing prices
have positive effects on aggregate consumption (Edelstein and Lum, 2004). The
private housing gains were, however, not translated into positive consumption in
Singapore (Phang, 2004).
The asymmetric wealth effects on household consumption may suggest that the two
housing markets are distinct and segmented. If household mobility occurs from the
public (starter) flats to the private (trade-up) houses, inter- and intra-market housing
price dynamics between two distinct submarkets on the property ladder can be
expected. The same hypothesis of housing price predictability as found in Case and
Shiller’s (1989, 1990) studies should also not be rejected between the housing
submarkets in Singapore.
3. The property ladder in the Singapore housing market: Empirical
Singapore has a two-tier housing market composed of a public market and a private
market. The public housing market provides new starter houses at subsidized rates for
first time home buyers, who meet specified social, demographic and income criteria. 5
There is a formal resale market for public housing units, where prices are market
The Central Provident Fund (CPF) is a comprehensive pension scheme, which has been expanded
to provide financing for public housing purchases in 1968, and subsequently for private housing
purchases in 1981 under the Residential Property Scheme (RPS).
See Phang (2004) and Bardhan, Datta, Edelstein and Lum (2003) for detailed discussions of policy
changes affecting the public housing market in Singapore over time.
determined. The difference in price between the new and resale public housing
markets creates a source of “fortuitous” wealth to qualified Singaporean first-time
home buyers (Bardhan, Datta, Edelstein and Lum, 2003). The public housing wealth
can only be cashed out by the households upon satisfying a 5-year minimum
occupancy requirement. The financial gains from the sale of the original public
housing units provide an important source of funding for down payments for liquidity
constrained households in trading-up for private (trade-up) houses.
The private housing market in Singapore operates on a laissez faire basis. Private
housing units are more expensive. They are also differentiated by better designs,
quality of finishes, and some (condominiums and apartments) are equipped with full
recreational facilities. Despite the relatively small scale of the private housing market,
housing types are more heterogeneous and range from landed units like detached
houses, semi-detached houses and terraces to non-landed units like condominiums
Based on the housing stock and average transaction prices summarized in Table 1, the
public and private housing markets could be arranged in a unique pyramid structure
(Figure 1). 6 The public housing market that constitutes nearly 80% of the total
housing stock forms the lowest stratum in the housing pyramid. This sub-market has
the lowest average transaction price of $2,923 per square meter (5-room flats). The
upper strata of the housing pyramid are made up of private housing markets that
constitute 20% of the total housing stock. The private housing market could be further
stratified into different submarkets based on the average transaction prices in an
ascending order starting from apartments, condominiums, terraces, semi-detached
houses and finally, detached houses. The hierarchical structure and distinct price
levels in the housing submarkets map well into the property ladder structure of
Ortalo-Magne and Rady (2004, 2006).
[Insert Table 1 and Figure 1]
A sandwiched class of housing, known as Executive condominiums (EC), was developed by the
HDB in mid 1996 to cater to young professionals who could not meet the income criteria for
public housing units. This segment of the market is excluded in this study.
Is the mobility of households on the housing ladder a strictly continuous process? If
liquidity constraints were strictly binding, household mobility is dependent on the
capital gains from the sale of their houses in the lower rung of the ladder. The gains
must be at least sufficient to make down payments for houses on the upper rung
immediately above the housing ladder. The condition implies that households do not
upgrade by moving more than one rung at a time in this model. 7 A starter flat (public
HDB unit) is the entry point of credit constrained households into the housing market.
The public housing owner “climsb up the housing ladder” by trading up a starter
home for an apartment or condominium, a terrace, a semi-detached house and finally
a detached house at the top of the pyramid in a sequential process. If this hypothesis is
not rejected, intra- and inter-stratum price discovery processes occur only between
two submarkets on contiguous rungs of the housing ladder. Price changes between
any stratum of the housing market and those not immediately above or below it are
expected to follow independent random walk processes.
The above independent random walk hypothesis seems to suggest that housing
submarkets on the housing ladder are segmented. However, if household mobility is a
continuous process, and if the upward and/or downward filtration takes place
simultaneously across different submarkets, price changes in one submarket may
trigger a series of reactions in other housing submarkets on the housing ladder. The
mobility and market segmentation hypotheses could be jointly tested by examining
price dynamics across different housing submarkets. If the upward mobility and the
integration of housing sub-markets hypotheses are both not rejected, long-run co-
movement of prices of different pairs of housing sub-markets in contiguous rungs on
the housing ladder could be expected.
Historical house prices in natural logarithm terms for the public and private sub-
markets in Singapore, both in levels and first differenced, are shown in Figures 2 and
3. Price changes in different housing sub-markets do not appear to follow random
walk processes. There are common long-term trends and turning points over the
sample period. More robust empirical tests will be carried out in the next section.
Although there are exceptional cases of households moving from lower rung starter houses to
private houses that are not immediately above the existing housing on the housing ladder, the
number of such cases is relatively small. The strict credit constraints condition will not be,
[Insert Figures 2 and 3]
4. Empirical Methodology
4.1. Model Specifications
The 1990s witnessed a highly volatile housing market in Singapore, which saw a
sharp rise in prices in 1994 to 1995, followed by a steep correction in 1996 and 1997
(Figure 2). 8 This paper applies the stochastic permanent breaks (STOPBREAK)
model proposed by Engle and Smith (1999) to empirically test the long-run co-
movement of housing prices. In the STOPBREAK model, stochastic shocks are
incorporated in the tests of long-run relationships between two time series. The
hypotheses of household mobility and market segmentation could be jointly tested in
the study by examining the price dynamics of different housing sub-markets. The
predictability of house prices using the house price information of other sub-markets
is also tested using a vector error correction model (VECM), which includes lagged
price changes and other fundamental economic variables. The two proposed modeling
strategies are discussed below.
4.1.1. Stochastic permanent breaks (STOPBREAK) model
In the classical cointegration methodologies proposed by Engle and Granger (1987)
and Johansen (1988), two d-th order integrated time series, I(d), are said to be
cointegrated, if the residual of the linear combination of the two series, [Yt = βXt +
ut], is stationary with an order of integration of less than d, [ut ∼I(<d)]. The definition
holds on the assumption that there are no short-term shocks to the equilibrium of the
system. If permanent and/or transitory economic shocks are stochastic, a pair of
variables may move together for a period of time and jump apart occasionally. Engle
and Smith (1999) call this process temporary cointegration, and they develop a
STOPBREAK model to capture the random structural shifts in the process.
Engle and Smith (1999) define the simplest form of STOPBREAK time series
process, yt as:
y t = mt + ε t , t=0,1,...,T (1)
The sharp reversal of housing prices from the boom occurred in 1997 and 1998 due to the
government’s intervention through a slew of off-budget anti-speculation measures in May 1996,
and as a result of the shocks from the Asian Financial Crisis in 1997 (Edelstein and Lum, 2004).
where mt = E[ yt I t −1 ] is a time varying conditional mean, and ε t is an error term.
mt = mt −1 + qt −1ε t −1 = m0 + ∑ qt −i ε t −i , t=1,2,...,T (2)
where qt = q(ε t ) ∈ (0,1) s.t E[qt ε t I t −1 ] = 0
Shocks in the STOPBREAK process are permanent and endogenously determined in
the process. If [ q =1], the realized process at time t is a random walk. If [ q = 0], the
conditional mean is constant, and the long-run forecast of yt will not be deviated from
the mean value of q . t
The general STOPBREAK process for a pair of time-series variables ( Yt , X t ) can be
extended as follows:
A( L) B( L)(Yt − X t δ ) = z t −1 A( L)ε t + (1 − z t −1 ) B ( L)ε t , t=1,2,...,T (3)
where [ A( L) = 1 − α 1 L − α 2 L2 − ... − α p L p ], [ B( L) = 1 − β 1 L − β 2 L2 − ... − β p L p ], L is
the lag operator, z t is a measurable function consisting of information up to t, and ε t
is an innovation term.
If [δ = 0], [B(L)=1-L], and [A(L)=1], the model complies with the simplest form of
the STOPBREAK process. Alternatively, if [δ ≠ 0], the two series are temporarily
cointegrated, where the two series jump apart occasionally and revert back to the
equilibrium relationship in the long-run.
To test the persistence of the STOPBREAK process, substitute [ qt (γ ) = ] into
γ + ε t2
processes (1) and (2) to derive the following specification:
− γε t −1
Δyt = + εt (4)
γ + ε t2−1
Based on equation (4), the random walk hypothesis, [Ho: γ = 0], is tested against the
alternative hypothesis, [H1: γ = γ ]. If the random walk hypothesis is rejected, the time
series contains a STOPBREAK process. One of the tests proposed by Engle & Smith
(1999) involves testing the null hypothesis [Ho: ϕ =0] against a negative alternative
using t-tests, where ϕ is estimated using the regression below:
Δy t −1
Δy t = ϕ + μt (5)
γ + Δy t2−1
The STOPBREAK methodology is applied to test the existence of temporary
cointegration in the long-term price processes of housing sub-markets. The tests are
conducted on a pair-wise basis for prices of public resale housing, apartment,
condominium, terrace, semi-detached and detached housing.
4.1.2. Vector Autoregressive Error Correction Model (VECM)
The mobility of households on a housing ladder suggests that some forms of the price
discovery process are likely to occur across different housing submarkets. Price
dynamics across different submarkets can be tested using a vector autoregressive
error-correction model (VECM), which consists of lagged price changes in housing
submarkets and other exogenous macro-economic variables. An error correction
mechanism is included in the model to correct for short-term deviations of the price
A generalized VECM system with n-lags can be specified as follows:
Δy t = δ i μ t −1 + ∑ α i Δy t −i + ∑ β i xt −i + ε t (6)
i =1 i =1
where Δyt is a vector of natural log-price changes for different housing submarkets
that include public resale HDB flats, apartments, condominiums, terraces, semi-
detached houses and detached houses, xt is a vector of exogenous variables that
include GDP growth, stock market return, prime lending rate, and unexpected
inflation rate, and εt is a vector of innovation terms. A time dummy variable,
Dum4Q98, which is given a value of 1 for periods in 4Q1998 and after, or zero
otherwise, is also included to capture the turning point that occurred in 1998. The
coefficient matrices αp and βp are jointly estimated in the VECM system with
appropriately selected p and q lags.
An error correction term, μt-1, in the VECM model represents the speed of adjustment
of the house prices toward the long-run values. If [δ ≠ 0], there exists a significant
cointegrating vector in the system, and the house price changes can be predicted in the
long-run. The hypotheses that household mobility triggers price adjustments from one
segment to another segment of the housing market, and that the housing submarkets
are not segmented, are both not rejected as a result.
4.2. Data Collection
House price time series data for the private housing sub-markets are collected from
the Real Estate Information System (REALIS) of the Urban Redevelopment Authority
(URA) of Singapore, and the public resale housing price data are obtained from the
HDB. The data series are published on a quarterly basis, which cover a 16-year
sample period from 1Q1990 to 1Q2006. Housing prices for different submarkets are
represented by the subscript i in the price variable, Pi, where [i = (det, semd, ter,
cond, apt, hdb)], and the corresponding natural log-terms of housing prices are
denoted by pi, where [pi = logn(Pi)]. Change in house prices is taken as the first order
difference of the natural log-price, [dpi,t = (pi,t – pi,t-1)].
We obtain time series data for four exogenous macro-economic variables, which
include GDP growth, stock market returns, the prime lending rate, and the consumer
price index from the Singapore Time-Series System of the Department of Statistics
(DOS), Singapore. The unexpected inflation is estimated from the consumer price
index using the interest rate model proposed by Fama and Gibbons (1984). The
notations for the variables and their respective sources are summarized in Table 2.
[Insert Table 2]
Augmented Dickey-Fuller and Phillips-Perron tests are conducted to test the order of
integration of the log-prices of different housing sub-markets and the macro-economic
variables. The results show that all housing price variables are I(1) stationary. For the
four fundamental variables, ADF test results show that the stock market return and
GDP changes are I(1), whereas unexpected inflation and the prime lending rate are
5. Empirical Results
5.1. Test of Temporary Cointegration in STOPBREAK process
The standard cointegration tests do not adjust for possible transitory and permanent
shocks in the time series. Two price series may move together and occasionally jump
apart as a result of an external shock. The random responses to shocks cause problems
to the cointegration tests, which may fail to detect long-run contemporaneous
relationships in price generating processes. In the Engle and Smith (1998)
STOPBREAK model with time-varying stochastic shocks, price series are said to be
temporary cointegrated if the random walk hypothesis is rejected. At least one
Granger causality relationship is expected between the price series, although they are
allowed to deviate apart occasionally following exogenous shocks.
The t-statistics for the STOPBREAK tests in Table 3 show that the random walk
hypothesis was rejected on a pair-wise basis between the HDB resale price and the
prices of all the five private housing submarkets. However, pair-wise temporary
cointegration relationships are not significant among the private housing submarkets.
The results imply that transitory or permanent shocks may cause the relative prices of
HDB resale units and other private housing sub-markets to drift apart in the short run.
These relative prices are, however, expected to converge to a long-run equilibrium
level. The relative prices of two housing submarkets will revert to fundamental levels
in the long-run, if the two housing price series were temporary cointegrated.
Therefore, according to Engle and Smith (1998), arbitrage could be made by selling
over-valued houses or buying under-valued houses, when random shocks occur.
New HDB units are sold by the government at concessionary prices to first-time
homebuyers who meet the HDB criteria. Increases in resale HDB price create
“fortuitous” gains to the HDB households, which can be converted into higher down
payments when households choose to move up the housing ladder. As HDB
households form the bulk of constrained upward movers, price changes in the HDB
resale market will significantly influence the rate of upward mobility of these
households. Policy changes that have direct impact on the public housing market
could also create “ripple” effects to the private housing market. Changes to public
housing finance policies through the Central Provident Fund (CPF) schemes in the
1980s and 1990s, and the relaxation of the HDB resale housing ownership rules in
1991 (Phang and Wong, 1997, Bardhan, Datta, Edelstein and Lum, 2003) were among
the significant shocks that may have caused permanent or transitory shifts to the
prices in the public and private markets. The temporary cointegration processes found
in the relative prices between public housing market and other private housing
submarkets in the STOPBREAK test are evidence of the policies’ impact on trading
up decisions of households.
[Insert Table 3]
5.2. Predictability of Housing Prices in VECM
In a vector autoregressive (VAR) system comprising all the housing prices, the
number of cointegrating vectors is first tested using the multivariate cointegration
methodology proposed by Johansen and Juselius (1990). Four fundamental variables:
dgdp, dsgxa, PLR and unexpinf, are included in the multivariate VAR model to
control for exogenous effects. The trace statistics reject the null hypothesis that there
are at most five cointegrating vectors, [r ≤ 5], at the 5% significance level. However,
in the Maximum Eigenvalue tests, the null hypothesis [r ≤ 2] was not rejected, which
implies that there exists at least one significant cointegrating vector in the VAR
For parsimonious considerations, one cointegrating equation is estimated in the
VECM for housing prices with two lags. A time dummy variable, DUM98Q4 is
included in the system to capture the break caused by the anti-speculation policy
shocks in May 1996 and the Asian financial crisis in 1997. The estimation results for
the VECM are summarized in Table 4.
[Insert Table 4]
The results show that the cointegrating vector, μt-1, was significant in explaining price
changes in all private housing submarkets, but not the HDB resale price changes.
Prices in private submarkets were predictable, and there was clear evidence of price
correction towards long term fundamental means in these markets.
The upward mobility of “starter” households to the upper rungs of the private housing
ladder could be reflected in the relationships between public and private house price
changes. The one-period lagged HDB resale housing price variable, dphdp,t-1, was
significant in explaining price changes in all private housing submarkets, whereas the
second period lagged term, dphdp,t-2, contains significant explanatory information on
prices of three private landed housing markets, dpter, dpsemd, and dpdet. All the lagged
public resale housing price variables have negative signs. There are two possible
reasons for the negative price relationships.
First, upward and downward housing mobility activities create short-term imbalances
in the supply and demand conditions in the public and private housing markets. The
action of HDB households in moving up the housing ladder brings about excess
demand in the private housing markets. At the same time, starter houses sold by the
constrained households cause excess supply in the resale HDB flats. Short-term
inelasticity in the supply of new private housing units creates disequilibrium in prices.
Prices of private houses escalate in the short run. Public housing price, on the other
hand, may decline or remain unchanged in the short-term when surplus stocks are
brought into the market by the trading-up households. The opposite effects are
observed when trading down activities occur. Secondly, public housing policies that
require HDB households to stay in their flats for a minimum period of 5 years before
selling their flats restrict their mobility in the short run. Public housing wealth is not
translated into new housing trading activities in the private housing markets in the
short term. The negative price relationships between the two housing markets may
reflect delayed mobility decisions by households in the public housing market.
The mobility behavior of constrained trade-up households and unconstrained repeat
households could be examined from the price dynamics for houses in the upper rungs
of the housing ladder. The results show that lagged condominium price, dpcond,t-1
exhibits significant positive price effects on the other four private housing sub-
markets. The signals of house price information in the private landed submarkets
were, however, mixed. The coefficients for lagged semi-detached house price
variables, dpsemd,t-1 and dpsemd,t-2, are positive and significant, whereas, the coefficient
for dpter,t-1, and dpdet,t-1 have negative signs in the landed housing price VECMs. The
positive price effects of semi-detached houses may reflect the investment motive of
the unconstrained households. The negative signs for the terrace coefficients could be
explained by the substitution effects among landed housing goods. Trading down by
unconstrained movers when prices decline as predicted by Stein’s model may offer an
alternative explanation to the negative price effects.
For the exogenous macroeconomic variables, the results show that stock market
coefficients are positive and significant in explaining price variations in
condominium, terrace and semi-detached housing markets. Prime lending rates were
found to have significant dampening effects on public resale flats and all private
landed houses. Unexpected inflation has negative effects on detached house prices,
and GDP growth dampens price changes in terrace houses. The negative effect of the
DUM98Q4 variable was only shown in public resale housing prices. The impact of
the post-1997 Asian financial crisis on private housing prices could have already been
reflected in the error correction process or other lagged price variables.
In summary, private housing prices were predictable by the error correction
mechanisms, μt-1, and the lagged public resale price variable, dphdb,t-1. These findings
together with the earlier results on temporary cointegration between public resale
housing and the five private housing submarkets suggest that the joint hypothesis that
household mobility takes place from the public to private housing markets, and that
the private housing submarkets are not segmented, could not be rejected.
When the liquidity constraint is binding in housing consumption, Stein (1995) and
Ortalo-Magne and Rady (2004 and 2006) predict that household mobility activities
create a positive relationship between house prices and transaction volumes.
Empirical studies thus far have focused on the impact of LTV constraints on housing
price and time to sale (Genesove and Mayer, 1997). The pricing side of the story in
the down-payment constraint model has not been empirically tested. Income shocks
that trigger households’ decision to move could be transmitted through a self-
reinforcing mechanism that runs from price, to down-payment, demand and then back
to price (Stein, 1995). Therefore, this study attempts to use evidence in housing price
dynamics to jointly test the price-volume and the market segmentation hypotheses. If
the hypotheses were not rejected, the price discovery process should flow efficiently
between different submarkets on the property ladder.
In the STOPBREAK tests, the results rejected the random walk null hypothesis for
HDB resale prices and the prices of all five private housing sub-markets. Exogenous
permanent or transitory shocks may cause the relative prices of the public resale
market and the private housing submarkets processes to drift apart occasionally, but
the prices are expected to converge towards equilibrium in the long run. In the VECM
estimations, the results show that the error correction terms are significant in
explaining the price variations in all five private housing sub-markets. Some degree of
predictability in the private housing submarkets is expected. The mobility of
households on a housing ladder triggers price effects across different housing
submarkets. This price information will only spill-over from one market to another on
the housing ladder, if they are not segmented.
Bardhan, A.D., Datta, R., Edelstein,R.H., Lum, S.K., 2003. A Tale of Two Sectors:
Upward Mobility and The Private Housing Market In Singapore. Journal of Housing
Economics 12, 83-105.
Case, K.E., Shiller, R.J., 1989. The Efficiency of the Market for Single Family
Homes. American Economic Review 79(1), 125-137.
Case, K.E., Shiller, R.J., 1990. Forecasting Prices and Excess Returns in the Housing
Market. AREUEA Journal 18(3), 253-273.
Case, K.E., Quigley, J.M., Shiller, R.J., 2005. Comparing Wealth Effects: The Stock
Market versus the Housing Market. Advances in Macroeconomics, 5(1), 1-32.
Edelstein, R.H., Lum, S.K., 2004. House Prices, Wealth Effects and the Singapore
Macroeconomy. Journal of Housing Economics 13, 342 – 367.
Engle, R.F., Granger, C.W.J., 1987. Cointegration and Error Correction
Representation, Estimation and Testing. Econometrica 25,1225-1228.
Engle, R.F., Smith, A.D., 1999. Stochastic Permanent Breaks. Review of Economics
and Statistics 81(4), 553-574.
Fama, E.F., Gibbons, M.R., 1984. A Comparison of Inflation Forecasts. Journal of
Monetary Economics 13, 327-348.
Genesove, D., Mayer, C.J., 1997. Equity and Time to Sale in the Real Estate Market.
American Economic Review 87(3), 255-269.
Johansen, S., 1988. Statistical Analysis of Cointegration Vectors. Journal of
Economic Dynamics and Control 12, 231-254
Johansen, S., Juselius, K., 1990. Maximum Likelikhood Estimation and Inference on
Cointegration – with Application to the Demand for Money. Oxford Bulletin
Economics and Statistics 52(2), 169-210.
Lee, N.J., Ong, S.E., 2005. Upward Mobility, House Price Volatility and Housing
Equity. Journal of Housing Economics 14, 127-146.
Lum, S.K., 2002. Market Fundamentals, Public Policy and Private Gain: House Price
Dynamics in Singapore. Journal of Property Research 19(2), 121-143.
Ortalo-Magne, F., Rady, S., 2004. Housing transactions and macroeconomic
fluctuations: a case study of England and Wales. Journal of Housing Economics 13,
Ortalo-Magne, F., Rady, S., 2006. Housing Market Dynamics: On the contribution of
Income shocks and credit constraints. Review of Economic Studies 73, 459-485.
Ong, S.E., Sing, T.F., 2002. Price Discovery between Private and Public Housing
Markets. Urban Studies 39(1), 57-67.
Phang, S.Y., 2004. House Prices and Aggregate Consumption: Do they Move
together? Evidence from Singapore. Journal of Housing Economics 13, 101-119.
Phang, S.Y., Wong, W.K., 1997. Government Policies and Private Housing Prices in
Singapore. Urban Studies 34(11), 1819-1829.
Slemrod, J., 1982. Down-Payment constraints: Tax Policy Effects in a Growing
Economy with Rental and Owner-Occupied Housing. Public Finance Quarterly 10,
Stein, J.C., 1995. Prices and Trading Volume in the Housing Market: A model with
Down-payment Effects. Quarterly Journal of Economics 110, 379-406.
Table 1: Characteristics of Private and Public Housing Markets in Singapore
Housing Type Average Average Average No of Housing % of
Floor/ Transaction Unit Caveat Stock (as Total
Land Area Price (S$)# Price Lodged at 4Q03)@ Housing
(sqm)# (S$ psm)# (as at Stock
A) Private Housing Market:
Detached house 1314.75 4,927,479 4,062 104 9,915 0.97%
Semi-Detached 340.10 1,440,098 4,452 143 20,628 2.01%
Terraced house 208.18 1,052,364 5,344 314 36,549 3.56%
Condominium 133.79 803,168 5,867 1513 85,869 8.36%
Apartment 125.46 743,830 6,274 801 57,973 5.65%
B) Public Housing Market:
HDB data: Total*3 815,633 79.45%
Executive flat 130 403,400 3,103 65,143 7.99%
5-Room flat 110 321,500 2,923 201,152 24.66%
4-Room flat 90 232,800 2,587 318,668 39.07%
3-Room flat 69 159,300 2,309 220,696 27.06%
# The statistics are computed based on caveats lodged on all private residential properties transacted
from 1 September 2003 to 31 December 20003, as captured in Real Estate Information System
(REALIS) of the URA.
@ The cumulative housing stock information is obtained from the time-series data in REALIS, URA
*1 The size of different public housing types is taken from typical new housing flat type built by HDB.
The size for different HDB flats may have changed over time.
*2 The resale price was the 4Q average resale HDB flats that were published on the Business time,
“Prices of HDB resale flats rise 1.2% in Q2 from Q1,” by Andrea Tan, 24 July 2004.
*3 The total housing stock number was obtained from the HDB annual report 2002/2003, and the
number does not include rental flats (53,141)
Note: Condominium and apartment are two common non-landed housing types built by private
developers, they are quite homogenous in physical, but prices differs especially between the
high- and middle-end condominium, and apartment. The total stock size of apartment is also
smaller, but the hierarchy structure is determined based mainly on price structure in this study.
Source: URA, HDB
Table 2: List of Variables and Derivations
Notation Variable Description Source
A) Housing price variables:
Phdb Resale HDB housing price index HDB
Papt Apartment price index URA
Pcond Condominium price index URA
Pter Terrace price index URA
Psemd Semi-detached housing price index URA
Pdet Detached housing price index URA
B) Macro-economic variables:
GDP Gross domestic products DOS1
SGX Singapore Exchange All-share index DOS
PLR Prime lending rate DOS
CPI Consumer price index DOS
unexpinf Unexpected inflation rate Estimated based on
Fama & Gibbons
C) First differenced variables: Derivation2:
dphdb ln-return of HDB price phdb,t – phdb,t-1
dpapt ln-return of apartment price papt,t -papt,t-1
dpcond ln-return of condominium price pcod,t – pcod,t-1
dpter ln-return of terrace price pter,t – pter,t-1
dpsemd ln-return of semi-detached price psed,t – psed,t-1
dpdet ln-return of detached price pdet,t – pdet,t-1
dgdp GDP Growth gdpt - gdpt-1
dsgx Stock market return sgxt - sgxt-1
1 DOS denotes Department of Statistics of Singapore.
2 The natural logarithm form of the variables are represented in lower case of the respective
Table 3: t-statistics for STOPBREAK Tests
Variable Apartment, Condominium, Terrace Semi- Detached
papt pcod house, pter Detached house, pdet
HDB resale 3.76* 4.01* 3.94* 3.62* 3.67*
Apartment, -1.16 -0.74 -0.18 0.78
Condominium, -0.66 0.34 1.53
Terrace house, -0.69 0.79
Ho: the relative price follows a random walk process.
* indicates significance at 5% level
Table 4: Vector Error Correction Estimates
dphdb dpapt dpcond dpter dpsemd dpdet
CointEq -0.0392 -0.3344 -0.3391 -0.2824 -0.3594 -0.5752
[-0.3671] [-4.7237] [-4.0304] [-3.3232] [-3.8103] [-6.7204]
dphdb(-1) 0.2823 -0.4790 -0.4434 -0.6234 -0.3895 -0.7309
[ 1.5447] [-3.9496] [-3.0762] [-4.2821] [-2.4102] [-4.9840]
dphdb(-2) -0.1065 -0.1538 -0.1491 -0.3719 -0.4572 -0.7311
[-0.4706] [-1.0242] [-0.8354] [-2.0630] [-2.2848] [-4.0264]
dpapt(-1) 0.1111 -0.2725 -0.1219 -0.0592 0.0079 0.0825
[ 0.5018] [-1.8539] [-0.6978] [-0.3357] [ 0.0404] [ 0.4641]
dpapt(-2) 0.3607 0.0208 0.0397 0.0872 0.0686 0.0575
[ 1.8260] [ 0.1588] [ 0.2551] [ 0.5543] [ 0.3928] [ 0.3626]
dpcond(-1) -0.0736 0.82195 0.6118 0.7426 0.8316 0.9518
[-0.2866] [ 4.8213] [ 3.0193] [ 3.6289] [ 3.6607] [ 4.6170]
dpcond(-2) -0.2694 0.4019 0.4108 0.1943 0.1150 0.1793
[-1.0610] [ 2.3850] [ 2.0510] [ 0.9608] [ 0.5120] [ 0.8798]
dpter(-1) 0.7475 -0.5474 -0.5254 -0.8804 -0.9858 -1.5260
[ 1.6156] [-1.7830] [-1.4399] [-2.3889] [-2.4098] [-4.1105]
dpter(-2) 0.2971 -0.1239 -0.2798 -0.3287 -0.3657 -0.8065
[ 0.9553] [-0.6004] [-1.1408] [-1.3266] [-1.3297] [-3.2319]
dpsemd(-1) -0.1617 0.4436 0.3426 0.8884 0.6803 1.3661
[-0.5531] [ 2.2873] [ 1.4864] [ 3.8159] [ 2.6324] [ 5.8254]
dpsemd(-2) -0.2453 0.4091 0.0758 0.6049 0.6687 1.1617
[-0.9637] [ 2.4220] [ 0.3773] [ 2.9833] [ 2.9709] [ 5.6875]
dpdet (-1) -0.2405 0.2315 0.3198 0.1429 0.2821 0.1630
[-1.2944] [ 1.8778] [ 2.1824] [ 0.9653] [ 1.7170] [ 1.0933]
dpdet(-2) -0.1568 -0.4653 -0.1658 -0.3575 -0.3546 -0.4441
[-0.8777] [-3.9246] [-1.1768] [-2.5124] [-2.2452] [-3.0983]
C 0.2505 0.0155 0.09513 0.2232 0.18606 0.3062
[ 3.2400] [ 0.3013] [ 1.5603] [ 3.6248] [ 2.7222] [ 4.9363]
dgdp -0.0589 -0.0748 -0.1659 -0.2668 -0.1059 -0.1756
[-0.4016] [-0.7684] [-1.4343] [-2.2844] [-0.8169] [-1.4927]
dsgx 0.0086 0.0790 0.0339 0.1130 0.0795 0.0621
[ 0.1864] [ 2.5901] [ 0.9336] [ 3.0847] [ 1.9556] [ 1.6830]
PLR -0.0348 -0.0033 -0.0156 -0.0344 -0.0296 -0.0491
[-3.1175] [-0.4457] [-1.7673] [-3.8738] [-2.9964] [-5.4899]
unexpinf 0.0026 -0.0089 -0.0412 -0.0449 -0.0417 -0.0688
[ 0.0694] [-0.3641] [-1.4148] [-1.5267] [-1.2779] [-2.3238]
DUM98Q4 -0.0573 0.0235 0.0045 -0.0262 -0.0102 -0.0164
[-2.9241] [ 1.8060] [ 0.2933] [-1.6766] [-0.5870] [-1.0402]
Adj. R- 0.5591 0.7740 0.6668 0.7070 0.7204 0.8431
Cointegrating Equation Estimates:
ContEq,μt-1 dphdb(-1) dpapt(-1) dpcond(-1) dpter(-1) dpsemd(-) dpdet(-1) C
Coefficient -0.7334 0.190551 2.031744 -4.5223 1.8059 1.0000 1.0825
t-statistics [-7.0064] [ 0.8676] [ 7.7727] [-8.3145] [ 4.4422]
Figure 1: Property Ladder in Singapore Housing Markets
Three groups of agents in Ortalo-
Magne and Rady (2005) and Stein
(1995) mobility models
Unconstrained repeat mover
Increase in Market Share (Stock)
Increase in Price (typical Unit)
Houses Constrained trade-up mover
starter flat buyer/ non-
Public Housing / HDB Flats mover
*Table 1 shows that the total housing stock of apartment is smaller than that of condominium. However, these two housing types are
homogenous, and are constantly grouped in the same non-landed private housing category. Based on the absolute price measure ,
apartment is placed in the right hierarchy in the housing pyramid/ladder.
Figure 2: Historical Prices of Public and Private Housing Markets in Singapore
Natural Log Price Index
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q
9 0 99 1 99 2 993 994 995 996 997 998 999 000 001 002 003 004 005 00 6
19 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2
Public (HDB) Apartment Condominium
Terrace Semi Detached Detached
Housing Price Changes (%)
Figure 3: Ex-post Price Changes of Public and Private Housing Markets