House price dynamics and their reaction

to macroeconomic changes
Article

Accepted Version

Nneji, O., Brooks, C. ORCID: https://orcid.org/0000-0002-

2668-1153 and Ward, C. W.R. (2013) House price dynamics
and their reaction to macroeconomic changes. Economic
Modelling, 32. pp. 172-178. ISSN 0264-9993 doi:
https://doi.org/10.1016/j.econmod.2013.02.007 Available at
https://centaur.reading.ac.uk/31493/

It is advisable to refer to the publisher’s version if you intend to cite from the
work. See Guidance on citing .

To link to this article DOI: http://dx.doi.org/10.1016/j.econmod.2013.02.007

Publisher: Elsevier

All outputs in CentAUR are protected by Intellectual Property Rights law,

including copyright law. Copyright and IPR is retained by the creators or other
copyright holders. Terms and conditions for use of this material are defined in
the End User Agreement .

www.reading.ac.uk/centaur

CentAUR
Central Archive at the University of Reading
Reading’s research outputs online
NOTICE: this is the author’s version of a work that was accepted for
publication in Economic Modelling. Changes resulting from the publishing
process, such as peer review, editing, corrections, structural formatting,
and other quality control mechanisms may not be reflected in this
document. Changes may have been made to this work since it was
submitted for publication. A definitive version was subsequently published
in Economic Modelling, 32 (2013), DOI: 10.1016/j.econmod.2013.02.007
House Price Dynamics and Their Reaction to Macroeconomic
Changes

Ogonna Nneji
ICMA Centre, University of Reading

Chris Brooks
ICMA Centre, University of Reading

Charles Ward
ICMA Centre, University of Reading

February 2013

Abstract

This article applies a three-regime Markov switching model to investigate the impact of the
macroeconomy on the dynamics of the residential real estate market in the US. Focusing on the
period between 1960 and 2011, the methodology implemented allows for a clearer understanding
of the drivers of the real estate market in “boom”, “steady-state” and “crash” regimes. Our
results show that the sensitivity of the real estate market to economic changes is regime-
dependent. The paper then proceeds to examine whether policymakers are able to influence a
regime switch away from the crash regime. We find that a decrease in interest rate spreads could
be an effective catalyst to precipitate such a change of state.

Key words: house prices, monetary policy, Markov switching model, macroeconomic drivers
JEL codes: G12, R31

Acknowledgement: the authors are grateful to two anonymous referees for their helpful
comments on a previous version of this paper.

Contact:
Chris Brooks
ICMA Centre
University of Reading
Whiteknights
Reading RG6 6BA,
UK
E-mail: [email protected]
Tel: +44 118 378 7809
Fax: +44 118 931 4741
 1. Introduction

Over the last few decades, residential property markets in many industrialised nations across the
globe have witnessed large cyclical variations in prices and volumes. Real estate cycles are often
characterised by a surge in prices followed by a fall or crash. An example of this was seen in the
UK housing market in the late 1980s. Financial liberalization in the UK led to a price boom, but
following an increase in interest rates, residential prices experienced a sharp decline in the early
1990s. More recently in the US, nationwide property prices grew by over 61% between 2000 and
2005 but fell sharply by 38% in the four years that followed. There are similar examples in other
countries including Japan, Ireland and Spain. These cycles are often linked with changes in
macroeconomic drivers such as interest rates and economic growth.

This cyclical nature of the residential real estate market has been a major topic of discussion over
the years mainly because a large proportion of the average household’s wealth is invested in
property. The housing market in the US accounts for more than 50% of the country’s fixed
capital stock (Baffoe-Bonnie, 1998). Economic theory suggests that wealth is one of the key
drivers of aggregate consumption in any economy, and so therefore, a downturn in the housing
market is likely to be followed by a decrease in household consumption levels, which may in turn
have adverse effects on the growth rate of an economy. Empirical evidence of this is shown in
Case et al. (2001), who use a panel of 14 countries in their study to present statistical evidence
that a 10% rise in housing wealth would lead to a 1.1% increase in consumption.

The behaviour of the residential real estate market is also important because of the impact of
house price falls on the lending portfolios of commercial banks and other financial institutions.
Wheelock (2006) shows that a period of large declines in house prices is very often followed by
an increase in the rate of mortgage defaults, which has an adverse effect on banks’ profits. The
reduction in profitability may lead to failures in banks and other real estate lenders and a
subsequent slowdown in economic activity. A recent high profile example of this was the
collapse of Lehman Brothers which was heavily exposed to the real estate market via mortgage
backed securities. Following the failure of Lehmans in September 2008, the CBOE Volatility
Index (VIX), often referred to as the “fear index”, jumped 70%,1 and a global recession followed
shortly thereafter. In their empirical study, Ghent and Owyang (2010) provide evidence that
house price changes drive business cycles. Thus, given the potentially detrimental effects of

1Whaley (2000) concludes that the VIX index is a gauge for investors’ fears as it provides an aggregate view
of their expectations of volatility in the stock market.

1
declining housing prices on the economy, a deeper understanding of the economic drivers of the
residential real estate market is required.

The purpose of this paper is to investigate how changes in key macroeconomic variables could
influence the growth in house prices, depending on which part of the cycle the real estate market
is in. This study examines the impact of macroeconomic drivers of real estate price changes in a
three-regime-switching context, thus providing information on how selected economic factors
influence price changes in the residential real estate market depending on whether the housing
market is a “boom”, “steady-state” or “crash” regime. The paper further contributes to the
existing literature by investigating the likelihood of monetary policy tools precipitating a switch
from a crash regime.

Most prior studies on this topic do not allow for significant structural breaks in house prices
resulting from huge upswings and collapses in prices, instead assuming that the relationship
between house prices and economic variables is stable and consistent. The sensitivity of house
prices to changes in these variables could, however, depend on the stage in the housing cycle.
Failing to account specifically for volatile periods such as the period between 2000 and 2012 in
the US housing market could produce results which may not reflect a true picture of the
relationship between macro factors and growth in house prices. Xiao (2007) and Nneji et al.
(2013a), studying the residential property market in Hong Kong and the US respectively, provide
evidence that structural breaks in real estate prices caused by speculative bubbles are likely to
disconnect the housing market from the cost of renting. As a result, the relationship between the
housing market and its macroeconomic determinants may be regime-varying, and so accounting
for the housing market cycle is critical when examining its response to external macroeconomic
factors. We seek to address several policy-relevant questions in this research. First, is the housing
market more sensitive to economic changes in boom or bust periods? Second, which economic
factor(s) has(have) the strongest impact on the housing market in each regime or cycle? Third, is
it possible for policymakers to influence a switch away from the “bust” state (typically
characterised by negative growth) using monetary policy tools?

Studying the US housing market between 1960 and 2011, we apply a three state Markov
switching model to examine the possibility that macroeconomic drivers of house price changes
are regime-specific. The regime-switching methodology implemented in this paper also enables
us to identify cycles in the housing market. We can then evaluate whether policymakers are able
to switch the housing market from a crash regime to a steady state simply by using monetary
policy tools. To our knowledge, this is the first study that applies a three Markov switching

2
model in the context of the relationship between the real estate market and the macroeconomy,
and the first study that examines the efficacy of policy tools for causing a switch away from the
“crash” regime.

The layout of the rest of the paper is as follows. Section 2 is dedicated to reviewing the existing
literature on the relationship between economic variables and the residential real estate market.
The following section, Section 3, provides descriptive statistics for the data used in the paper. In
Section 4 a more detailed explanation of the methodologies implemented is given. Section 5
interprets and discusses the results from the Markov switching regressions, and in Section 6 we
examine the power of monetary policy tools in influencing a switch in the housing market from a
crash regime. We conclude in Section 7.

2. Literature Review

The literature on the macroeconomic determinants of house prices is vast. For many years,
especially before the subprime mortgage market crisis of the late 2000s, researchers used linear
models to examine the relationship between the macroeconomy and the dynamics of house
prices. These papers have generally applied regression analyses to evaluate how growth in real
estate prices is being driven by economic factors including interest rates, inflation, the
unemployment rate, economic growth.

The majority of these papers identify interest rates as the most important explanatory variable.
One early study is Abraham and Hendershott (1992). Using pooled cross-sectional data on
metropolitan house prices in the US between 1977 and 1991, they find that macroeconomic
factors including interest rates and employment are significant in influencing house prices.
Iacoviello and Minetti (2003) argue that, over time, house prices became more sensitive to
interest rate changes due to financial liberalization in European countries including the UK.
More recently, this sentiment is shared by Himmelberg et al. (2005), who show the importance
of interest rates to house price changes. The authors also find that the sensitivity of house prices
to long-term interest rates intensifies when rates had been relatively low in the recent past.
Furthermore, they cite that house prices are even more sensitive to long-term interest rates in
cities where values grow relatively faster. Adams and Füss (2010) distinguish the impact of short
term from long term interest rates on real estate price dynamics. They argue that short term
interest rates adversely affect demand for houses mainly because of the effect on mortgage rates
and the cost of financing for construction firms. Holly and Jones (1997), McQuinn and O’Reilly

3
(2007) and Bouchouicha and Ftiti (2012) have also investigated the link between the real estate
market and interest rates.

Other papers have established that price dynamics in the real estate market may be influenced by
macroeconomic factors other than interest rates. For example, Lastrapes (2002), using a vector
autoregressive model, attributes short run increases in house prices to positive money shocks.
Brunnermeier and Julliard (2008) conclude that inflation, not real interest rate changes,
influences the price-rent ratio as it is a signal of a likely future downturn in the economy.
Investigating the linkage between macroeconomic fluctuations and international house prices,
Beltratti and Morana (2009) estimate that 40% of the variation in house prices in G-7 countries
is caused by global macroeconomic shocks. The study by Adams and Füss (2010) provides
evidence that variables linked with economic activity such as industrial production, the level of
unemployment and money supply influence demand for housing and house price. Other papers
that study the macroeconomy-housing relationship on an international scale include Englund
and Iaonnides (1997), Tsatsaronis and Zhu (2004) and Glindro et al. (2011).

As discussed above, only a handful of studies have accounted for the housing cycle when
examining the relationship between house prices and the macroeconomy. A notable exception is
the research by Hall et al. (1997), who use a two-regime switching error-correction approach to
develop a macroeconomic model for UK house prices which accounts for booms and busts.
Their paper uses a Markov regime-switching approach, assuming that there are only two distinct
regimes in the housing market – “boom” and “bust”. The present paper adds a third regime
which we call “steady-state” based on the premise that the US housing market may have been in
a dormant state where prices are growing at a small steady rate. In this state, the housing market
is neither booming nor declining in prices. This can be observed in the early 1990s where
nominal house prices grew by an average of just 3% per annum in contrast to the early 2000s
boom of 11% annual growth and the late 2000s bust which saw prices decline by an average of
6% per year. Furthermore, this study builds on the work by Hall et al. as we model the
government’s ability to affect a regime shift, which thus has policy implications.

3. Data

The data employed in this study consist of sets of 203 quarterly observations from 1960Q1 to
2011Q3. For residential property market prices, the data are supplied by the Lincoln Institute of
Land Policy, constructed using the methodology of Davis et al. (2008). In order to compute this

4
series, Davis et al. use interpolation procedures to adjust the quarterly changes in the Freddie
Mac Conventional Mortgage House Price Index (CMHPI) in order to connect the series with the
Decennial Census of Housing benchmark levels. The price series used in this study is adjusted
for inflation.

In this paper, we employ inflation, disposable income growth, the short rate and the term
structure of interest rates (also referred to as the yield curve or interest rate spread) as key
economic variables that influence housing demand. These variables have been selected because
of their significant contribution to the dynamics of house prices suggested in the literature. For
example, studying the effect of economic risk factors on growth in the US commercial real estate
market, Ling and Naranjo (1997) identify unexpected inflation, yield spread and the Treasury bill
rate as its key drivers. Similarly, the study by Brooks and Tsolacos (1999) shows that the term
structure and unexpected inflation explain changes in the UK property market. Englund and
Ioannides (1997) find that lagged interest rate and influences house price dynamics in 15 OECD
countries. Using a life-cycle model of the property market, Ortalo-Magne and Rady (2006) show
that house price dynamics are driven by disposable income.

These drivers are likely to have the following impacts on real estate prices. An increase in the rate
of interest is expected to drive borrowing rates up, thus increasing the cost of servicing
mortgages. This would, in turn, lead to a decrease in demand for properties and a subsequent fall
in prices. Rising interest rates might additionally be expected to lead to an increase in the number
of defaults on mortgages (especially adjustable-rate mortgages), which could also lead to a fall in
prices. An increase in the term structure means that the gap between long and short term rates
widens. When this happens, market agents would expect future increases in the short term
interest rate and so house prices may fall. In addition, most housing finance is arranged at the
long maturity end, and thus upward twists in the yield curve will also lead to rises in borrowing
costs for mortgage holders. Growth in disposable income is expected to have a positive impact
on the housing market whilst the effect of inflation on the market should be negative.

The explanatory variables are constructed as follows. The short term rate used is the yield on 3-
month US Treasury bills. The disposable personal income measure used in the paper is obtained
from the US Bureau of Economic Analysis. It is the proportion of income available to
individuals for spending after deducting personal current taxes. The term structure is constructed
by taking the difference between the long rate and the short rate. We take the constant maturity
rate on 10-year Treasuries as a proxy for the long interest rate. The inflation rate is computed
from the Consumer Price Index (CPI) series. The economic data series are retrieved from the

5
online database of the Federal Reserve Bank of St. Louis.2 A set of summary statistics for our
sample is presented in Table 1.

Table 1: Summary Statistics

Here, HP, CPI, INC, LTR, STR and TS are acronyms for house prices, the consumer price index, disposable
income, long term interest rates, short term interest rates and the term structure, respectively. The symbol “Δ”
represents the percentage changes in the series. Note that the data series for house prices and interest rates are
adjusted for inflation.
Δ HP Δ CPI Δ INC LTR STR TS
Mean 0.29% 0.98% 0.54% 5.67% 4.72% 0.95%
Maximum 4.53% 3.80% 4.33% 13.40% 13.54% 3.35%
Minimum -8.30% -2.38% -2.55% 1.67% -0.88% -1.94%
Skewness -1.86 0.67 -0.02 1.14 0.67 0.05
Kurtosis 8.95 5.98 5.23 4.16 4.15 2.58

The descriptive statistics on house prices, CPI and INC are the percentage changes of these
series, thus representing the house price growth rate, inflation rate and disposable income
growth rate, respectively. From the table, it is clear that the average real growth in house prices is
a modest 0.29% per quarter or 1.17% per annum. Prices have grown as much as 4.53% during
the boom years, whilst in market downturns such as the recent crash, price falls have been more
severe (-8.30%). The distribution of returns on investing in the housing market is negatively
skewed, thus implying that it has a longer left tail. Another key factor to note is that the
distribution of house price growth is leptokurtic (i.e. kurtosis greater than 3), which implies that
the data are highly concentrated around the mean. The second column refers to the percentage
changes in CPI, i.e. the inflation rate. From our sample, the average quarterly rate of inflation is
0.98%, which equates to an annual inflation rate of almost 4%. The real growth in disposable
income, denoted by the percentage change in INC, has an average of 0.54% per quarter or
2.16% per year. The greatest drop in income in our sample set is -2.55%, occurring at the end of
2008, which was fuelled by the economic decline caused by the “meltdown” of the subprime
mortgage market. The early 1980s saw the highest short and long term interest rates, 13.40% and
13.54%, respectively. As for the term structure, it has an average of 0.95% and a maximum of
3.35%.

4. Methodology

2For more information on the original sources of the data, see the website of the Federal Reserve Bank of St.
Louis: http://www.research.stlouisfed.org/

6
As stated earlier in this paper, most of the studies that have attempted to examine the
relationship between economic variables and the real estate market (either commercial or
residential) have used linear regression type models. For example, Levin and Wright (1997) use a
pooled cross-sectional regression approach to regress house prices on several explanatory
variables including real interest rates while Hort (1998) applies an error correction model to
study the impact of demand and supply-based market fundamentals on house prices in Sweden.
Other examples include the papers reviewed in Section 2.

Building on from these studies, in this paper, we apply a three-regime univariate Markov
switching (MS) model due to Hamilton (1989, 1994). This methodology allows us to study the
nonlinear relationship between the growth in house prices and changes to economic
fundamentals:

HPt  St ,0  St1 ,1STRt 1  St1 ,2 TSt 1  St1 ,3CPIt 1  St1 ,4 INCt 1   St (1)

 
where  St ,t ~ N 0,  S2t and St  i (for i = 1, 2 or 3)

The estimated betas from model (1) can be interpreted as a measure of the sensitivity of house
price growth to changes in the interest rate, term structure, inflation and disposable income.
Notice that these explanatory variables are lagged and not contemporaneous because, unlike
other asset classes such as the stock market, changes in the economy are expected to have a
delayed effect on the housing market. This formulation also avoids the possible concern that
there may potentially be an endogeneity problem if there are feedbacks from the housing market
to the macroeconomy or if house price dynamics affect monetary policy. We assume the lag time
to be one quarter.

Our model is a three-regime MS model, where the term St is the latent state variable which could
take the value of 1, 2 or 3 depending on the state or regime in the housing market. In other
words, the effect of each of the explanatory economic variables depends on the housing cycle i.e.
whether there is a boom, bust or steady growth in the housing market. This unobservable state
variable is governed by a first order Markov chain with a constant transition probability matrix
(P):

 Pr  St  1| St 1  1 Pr  St  2 | St 1  1 Pr  St  3 | St 1  1   p11 p12 p13 

 
   Pr  St  1| St 1  2  Pr  St  2 | St 1  2  Pr  St  3 | St 1  2     p21 p22 p23  (2)
 Pr  St  1| St 1  3 Pr  St  2 | St 1  3 Pr  St  3 | St 1  3   p31 p32 p33 

where pij are the transition probabilities from state i to state j.

7
The MS model is estimated using a maximum likelihood procedure. Under the assumption that
the error term   t  is normally distributed, the density of yt conditional on the regime  i  is

represented as:

  yt  X t1i  
 2
1 
i ,t  f  yt | St  i, X t , t 1;   exp   (3)
2 i
2

 2 2
i 


where, t 1   yt 1 , yt 2 ,..., X t 1 , X t 2 ,... represents all the past information to time t–1,  is the


vector of parameters  St ,0 ,  St ,1 ,  St ,2 ,  St ,3 ,  St ,4 ,  S2t , p00 , p11 , p22  to be estimated. Note that

yt is the dependent variable while X t is the vector of explanatory variables. The conditional

density of the observation at time t is obtained from the joint density of yt and St :

f  yt | t 1;   f  yt , St  1| t 1;   f  yt , St  2 | t 1;   f  yt , St  3 | t 1;  (4)

which is equivalent to:

3

 f y |S
i 1
t t  i, t 1; P  St  i | t 1;  (5)

As it is impossible to know for certain what regime the housing market is in, inference about the
regime is made by observing the growth rate of house prices. The inference comes in the form
of filtered probabilities  jt  , which are computed recursively using historical information, t 1 :

p  jt
ij i ,t 1
 jt  Pr  St  j | t ;   i 1
(6)
f  yt | t 1 ; 

These filtered probabilities depend on real-time updated information up to time t. It is also

interesting to compute probabilities of what state/regime the housing market was in at a
previous date t using all the observations and information obtained through a later date T. These
are known as smoothed probabilities it|T  Pr  St  i | T ;   , and they are computed using an

algorithm developed by Kim (1994). Estimation of the parameters  in the MS model is done by
maximizing the following log-likelihood function:
T
l     f  yt | t 1;  (7)
t 1

8
 5. Results

Before applying the three-regime MS model, we examine the same regression in equation (1) for
the whole period using a standard linear ordinary least squares (OLS) estimation procedure. In
this OLS model, the subscript St is omitted from the right hand side variables of the equation.
Note that the estimated coefficients are standardised following the approach introduced by Bring
(1994). This allows us to determine the relative importance of the selected explanatory variable,
thereby making it easier to interpret the output from the OLS model. The results are shown in
Table 2.

Table 2: OLS Model Output

We present a linear regression model of equation (1) for a single regime using ordinary least squares estimation
procedure. This time series regression analysis is performed on the whole sample between 1960 and 2011. The
dependent variable is the quarterly change in house prices and the independent variables are the changes in the
short term interest rate (ΔSR), the term spread (ΔTS), inflation rates (ΔCPI), and disposable income (ΔINC). The
figures in parentheses are the p-values and the grey shade indicates variables that are statistically significant (10%
at least).
Constant Δ SR Δ TS Δ CPI Δ INC
-0.0020 -0.2283 -0.1935 -0.0389 0.1645
β
(0.35) (0.01) (0.03) (0.58) (0.02)

In this linear single state model, the sign of the short term interest rate is negative and it is
statistically significant, implying that an increase in the rate of interest would have a negative
effect on house prices, as expected. The term structure is also negative as anticipated and it is a
statistically significant driver of house price dynamics. The inflation rate parameter estimate is
positive but statistically insignificant. Growth in disposable income, however, does have a
relatively large and positive beta, which implies that an increase in disposable income is expected
to have the most significant positive effect on house prices of all the economic variables
investigated. These results raise crucial questions, such as whether it is really true that the
inflation rate has no effect on real house prices, and also, whether it is the case that the sensitivity
of the housing market to economic changes remains constant between different regimes.
Implementing the three-regime MS model provides answers to these questions.

Table 3: Determining the appropriate number of regimes using information criteria

In this table, we compare the Akaike (AIC), Hannan-Quinn (HQIC) and Schwartz (SIC) information criterion to
determine the optimal number of regimes to be used to model house price changes.

9
Information Criterion 1 Regime 2 Regimes 3 Regimes 4 Regimes

AIC -5.2731 -6.1473 -6.4518 -6.4652

HQIC -5.2403 -6.0424 -6.2748 -6.2029

SIC -5.1920 -5.8880 -6.0142 -6.0142

Note that in order to prove that the statistically appropriate number of regimes is three, we
observe and compare the Hannan-Quinn, Akaike and Schwarz information criteria of two-,
three- and four-regime versions of the MS model. A two-regime model would imply that there is
just a “boom” and “bust” state while a four-regime model would signify a “crash”, “boom”,
“slow growth” and “recovery” (Ryden et al., 1998 and Guidolin and Timmermann, 2008). In our
case, we find that the Hannan-Quinn and Schwarz information criteria using all the different
regime approaches are smallest in the three-regime Markov Switching model, thus we infer that
the optimal number of regimes to be used is three: “boom”, “steady-state” and “bust”. This is
shown in Table 3. AIC selects four regimes, but the criterion value is only slightly higher for the
three-regime model; we consider four regimes to represent an over-parameterisation of the data,
and therefore we implement the three-regime approach. The table also highlights the inadequacy
of the linear model specification, since the values of all three information criteria are far inferior
for the 1-regime (i.e. linear) model than for any of the switching models.

Prior to discussing the estimated regime-dependent betas, it is important to view and discuss the
estimated smoothed probabilities of the housing market being in any of these growth regimes.
These probabilities allow us to make statistical inferences about the regime in which the housing
market resides at each point in time by observing the complete dataset. Recall, the smoothed
probabilities are dependent on the estimated transition probability matrix which provides
information on the probability of a switch from one state at time t - 1 to another at time t as
depicted in equation (2). The estimated transition probability matrix is shown in equation (8):

 p11 p12 p13   0.953 0.025 0.039

   p21 p22 p23   0.047 0.951 0.000 (8)
 p31 p32 p33   0.000 0.024 0.960

where 1 = Steady-State, 2 = Boom and 3 = Crash regime.

The probability of remaining in the crash regime given that the housing market was in the crash
regime in the previous period is 96%, meaning that this is the most persistent regime in the
housing market over the last 50 years. There is a 2% chance of switching from the steady-state

10
regime to the boom regime, whilst there is a 3.9% chance of a transition to a crash regime when
prices are growing steadily. The boom regime has the least persistence in that there is a 95.1%
probability of remaining in the boom period if there was a boom in the last quarter. With these
transition probabilities, it is also possible to compute the expected duration (ED) of being in
each of the regimes. This is simply calculated using the following formula:

ED  1 (9)
1  pii 
Therefore, the expected duration of being in the steady-state, boom and crash regimes are
roughly 21, 20 and 25 quarters, respectively. This means that we would expect a steady growth
and housing boom regimes to last for five years and a housing bust last for just over six years. A
graphical representation of the smoothed probabilities is given in Figure 1.

Figure 1: House price dynamics and regimes

6% 1.0

4%
0.8
2%
House price growth

Probabilities
0%
0.6
1960

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

2005

2008

2011
-2%

0.4
-4%

-6%
0.2
-8%

-10% 0.0
House Price Growth Steady-State Boom Bust

From Figure 1, we note that the dominant state is as expected steady-state although the stability
of the market is quite clearly disturbed by the policy changes that took place in the 2000s. It is
clear that in the 1960s, real house prices grew at a steady pace of below 2% per quarter, until the
market witnessed a dip. This dip coincided with the 11 month recession of the late 1960s as
noted by the National Bureau of Economic Research. However, the housing market soon
returned to its steady state and then a boom as prices picked up in the early 1970s. This lasted

11
until towards the end of the decade, which saw a transition back to the steady-state regime and
house price growth slowed. From then on, house prices continued to grow steadily, averaging
just over 1% per quarter until the turn of the decade. Following the collapse of the tech bubble
at the end of the 1990s, policymakers reacted by adopting expansionary monetary actions. This
involved severe interest rate cuts, which made borrowing a lot cheaper than in previous years.
Furthermore, deregulation meant that the securitization market grew significantly, making it
easier for borrowers to obtain mortgages. House prices began rising faster than they had done at
any time in the previous 40 years and there was a transition to the boom regime. Several papers
including Shiller (2007), Coleman et al. (2008) and Nneji et al. (2013b), attribute this period’s
growth in prices to the existence of a speculative bubble in the market. A crash then followed
this during 2000 – 2005, which saw the biggest drop in house prices in recent history, and there
was a meltdown in the mortgage market. Although there were two consecutive quarters in 2009
which saw real house prices grow by almost 5% combined, further dips occurred and the market
soon reverted back to the crash regime. Table 4 provides the estimated parameters of the MS
model which aims to provide a detailed insight into how these four economic variables influence
the growth in house prices in the three regimes.

Table 4: Markov Switching Model Output

. This time series regression analysis is performed on the whole sample between 1960 and 2011. The dependent
variable is the quarterly change in house prices and the independent variables are the changes in the short term
interest rate (ΔSR), the term spread (ΔTS), inflation rates (ΔCPI), and disposable income (ΔINC). The figures in
parentheses are the p-values and the grey shade indicates variables that are statistically significant (at the 10%
level or better).
Regime Constant Δ SR Δ TS Δ CPI Δ INC

0.0037 -0.1165 -0.0053 -0.1431 0.1280

Steady-State
(0.01) (0.00) (0.90) (0.00) (0.01)
0.019 -0.1025 -0.0928 -0.2491 0.0028
Boom
(0.00) (0.15) (0.08) (0.00) (0.96)
-0.0204 -0.3961 -0.2226 -0.1785 0.2686
Bust
(0.01) (0.30) (0.64) (0.68) (0.54)

From Table 4, there are clear differences in the signs, significances and sizes of the estimated
betas depending on the regime. Note that the estimated coefficients are also standardised in
order to enable us to easily compare the relative importance of each of the explanatory variables.
For example, changes in the short term interest rate influence the housing market in the steady-
state and boom regimes only. Focusing on the size of the coefficients, it is clear that interest rate
cuts have a more severe effect on the growth of house prices in the steady-state regime than in
any other regime. Similar to our finding from the linear model that the term spread influences
growth of house prices, in the boom regime, an increase in the term spread or a widening of the

12
gap between long and short term interest rates is expected to cause a fall in the growth rate of
house prices, perhaps as a result of investors’ expectations of future increases in the cost of
borrowing. In the steady-state and crash regimes, the term spread has no effect on house price
dynamics. However, unlike the linear model which depicted that the rate of inflation was
statistically insignificant, we find that inflation does in fact have an effect on the growth of house
prices in the boom and steady-state regimes. The last driving variable to be examined is
disposable income growth. In the MS model, percentage changes in disposable income positively
influence growth in house prices in the steady-state regime only.

6. Is monetary policy a catalyst in instigating a switch from the crash regime?

It is worth mentioning that the housing market becomes completely insensitive to changes in the
economic variables in the crash regime; none of the variables are seen to have any effect on
house price changes. Given these findings, it is clear that monetary variables affect the real estate
market within a regime, but could policymakers instigate a transition from the crash regime to
either the steady-state or a boom regime? In order to answer this question, we employ a probit
model. We observe the filtered probabilities generated from the estimation of the MS model,
focusing on the probability of being in the crash regime. The probit model is a binary
specification, and thus we assign 1 (y =1) if the observed filtered probability (y*) of being in the
crash regime is greater than 0.33 (as there are three regimes) and 0 otherwise – i.e., the left hand
variable of the probit model, y, equals 1 if the probability of being in the bust regime is greater
than 33%. This allows us to examine whether monetary policy tools such as the short term
interest rate, the term spread and the level of money supply (M2) could be used by policymakers
to instigate a switch away from the crash regime. We also include a macroprudential tool
available to policymakers – the loan-to-income ratio. Here, “loan” refers to the home mortgage
debt outstanding in the economy. This is a leverage proxy used in measuring how easy it is to
obtain credit, whereby a high ratio represents easy credit conditions. Thus in this section, we
measure how alterations to these policy tools could reduce the probability of being in the crash
regime. The probit model is as follows:

Pr  y*  0.33  Pr  y  1| x   F  0  1dummyLTI t 1   2dummySRt 1  3dummyMSt 1   4dummyTSt 1 

(10)

where dummyLTI, dummySR, dummyMS and dummyTS refer to dummy variables which assign 1
when there are increases in the loan-to-income ratio, short term interest rate, money supply or
term spread, respectively and 0 when there are no changes or decreases occur.

13
In this probit model, F(.) follows a cumulative normal distribution:
x
F  x    x     z dz (11)


where the normal density function, (z), is given by:

 z2 
exp   
  z   2 (12)
2

A statistically significant positive αi (i=1,2,3) would imply that an increase in the value of one of
these monetary tools would increase the probability of remaining in the crash regime in the next
quarter whilst a negative αi that is statistically significant implies the reverse. This part of the
study provides information as to whether monetary policy approaches may be suitable for
periods of severe downturns in the housing market. The results of the probit regressions are
reported in Table 5.

Table 5: Probit Model Output (Including LTI)

We present the probit model parameter estimates as stated in equation (9). The figures in parentheses are p-values
and the grey shade indicates variables that are statistically significant (at the 10% level or better). The McFadden and
Estrella R-squared figures for these probit maximum likelihood estimates are 0.03 and 0.02, respectively.
Constant dummyLTI dummySR dummy TS dummyMS
-1.1281 -0.0998 0.1410 0.4212 -0.2015
α
(0.00) (0.65) (0.55) (0.07) (0.39)

It is immediately obvious from Table 5 that, apart from the intercept, only one variable, the term
spread, has a statistically significant parameter estimate. This implies that changes to short term
interest rate levels and the money supply would have no effect on forcing the housing market
away from the crash regime. Therefore, in times of significant crashes in the housing market as
recently witnessed, money supply increases or short term interest rate cuts would not be
sufficient to encourage a switch away from this regime. The coefficient representing the term
spread, on the other hand, is positive and statistically significant, thus implying that an increase in
the term spread would increase the probability of remaining in the crash regime. Therefore, a
decrease in the spread would reduce the probability of being in the crash regime, thereby
increasing the probability of switching to either a boom or a steady-state regime. This means that
the interest rate spread may contain relevant forward-looking information about the future
housing cycle beyond the information contained in the money supply and short term interest
rates. Our finding is consistent with those in the economic literature postulating the effectiveness

14
of the yield spread in predicting future economic activities and business cycles. Many of these
papers have found the interest rate spread to be the leading indicator of business cycles. For
example, Estrella and Hardouvelis (1991) demonstrate the usefulness of the spread between the
10-year Treasury bond rate and the 3-month T-bill yield in predicting the probability of a
recession in the US since 1950. Also, Haubrich and Dombrosky (1996) find that the term spread
is an efficient forecaster of economic growth up to four quarters ahead. Other papers to have
investigated the predictive content of yield spreads in forecasting business cycles include Dueker
(1997), Estrella and Mishkin (1998), Dotsey (1998) and Hamillton and Kim (2000).

Policymakers may rely on other housing-specific non-monetary policy tools in dealing with real
estate busts. Crowe et al. (2011) evaluate the effect of macroprudential regulations and fiscal as
well as monetary policies in dealing with real estate market booms rather than busts. They note
that monetary policy changes may not be the most efficient way to influence the real estate
market because these changes would affect other relevant economic factors such as the level of
unemployment. Their study highlights possibilities that fiscal policy approaches, such as
alterations to taxes on homeownership, could be effective in taming exuberance in the property
market. They argue that other fiscal tools could include changes to housing transactions tax rates
and mortgage interest tax deductibility. Crowe et al. also show that macroprudential rules, such
as strict limits on loan-to-value ratios, also work in reducing house price booms. Given their
findings, it may be possible that prudent property tax rate cuts and increased loan-to-value ratios
may help the real estate market move away from the crash regime without having a significantly
huge effect on other aspect of the economy. It would also be of interest to investigate the
influence of fiscal tools on the regime transitions, but the lack of availability of quarterly
information dating back to 1960 makes it impossible to statistically test this.

7. Conclusions

In this paper, we employ a three-state Markov switching nonlinear econometric model to

examine the relationship between the residential real estate market and key macroeconomic
variables in the US. We investigate the regime-dependent effects of changes to short term
interest rates, term spread, inflation and GDP on house prices – i.e., we study the effect of these
macroeconomic variables on house price dynamics in times of housing booms, busts and
tranquility. The paper then proceeds to evaluate the effect of policy changes on switching
regimes in the housing market. We find the following. First, there is statistical evidence to
suggest there to be three distinct regimes in the housing market, namely “steady-state”, “boom”

15
and “crash” regimes. Changes to the selected macroeconomic variables significantly affect the
dynamics of house prices in the steady-state and boom regimes only. During housing busts,
however, the real estate market disconnects from these macroeconomic fundamentals. We also
find evidence of varying degrees of sensitivity of house prices to changes in these economic
variables, with prices generally being more sensitive during housing booms. We also show that
using a standard, single-state methodology inappropriately indicates that house prices are
responsive to changes in the short term interest rate and GDP only, thus providing an
incomplete view of the market.

Our second set of key findings in this study is based on the estimated probabilities of switching
from one regime to another. We find that there is a 5% chance of moving from a housing boom
in one quarter to a housing bust in the next quarter, but the most persistent regime is the steady-
state as there is a 98% chance of remaining within it if the housing market was previously in the
steady-state. Third, we show that monetary policy could potentially be used as a tool to enforce a
switch away from housing busts. Using a probit model based on the estimated filtered
probabilities from the Markov switching model, we find there is a statistically significant and
positive relationship between increases in the term spread and the probability of being in the
housing crash regime, thus implying that a reduction in the spread between long and short term
interest rates reduces the probability of being in the crash regime. Interest rates and money
supply cannot, however, be used to instigate a switch away from a housing bust.

Although this study provides a detailed examination of the impact of macroeconomic

fluctuations on the housing market, we do not evaluate the effect of other non-monetary policy
changes on the market due to data constraints, and thus further research could be conducted in
this area. If researchers were able to access unique fiscal data, they could extend this study to
investigate whether it is possible to use fiscal policies to enforce a transition away from housing
busts. This would allow policymakers to judiciously select the most appropriate tools to use in
reinvigorating a collapsing real estate market or possibly dampening demand during times of
overheating.

16
REFERENCES

Abraham, J.M. and Hendershott, P.H. 1992. Patterns and determinants of metropolitan house
prices, 1977 to 1991, In Lynn E. Browne and Eric S. Rosengren, (eds.), Real Estate and the Credit
Crunch, Boston, M.A: Federal Reserve Bank of Boston, 18 – 42.

Adams, Z., and Füss, R. 2010. Macroeconomic determinants of international housing markets,
Journal of Housing Economics, 19, 38 – 50.

Baffoe-Bonnie, J. 1998. The dynamic impact of macroeconomic aggregates on housing prices

and stock of houses: A national and regional analysis, Journal of Real Estate Finance and Economics,
17, 179 – 97.

Beltratti, A. and Morana, C. 2009. International house prices and macroeconomic fluctuations,
Journal of Banking and Finance, 34, 533 – 45.

Bouchouicha, R. and Ftiti, Z. 2012. Real estate markets and the macroeconomy: A dynamic
coherence framework, Economic Modelling, 29, 1820 – 29.

Bring, J. 1994. How to standardize regression coefficients, The American Statistician, 48, 209-13.

Brooks C and Tsolacos, S. 1999. The impact of economic and financial factors on UK property
performance, Journal of Property Research, 16, 139 – 52.

Brunnermeier, M.K. and Julliard, C. 2008. Money illusion and housing frenzies Review of Financial
Studies, 21, 135 – 80.

Case, K., Quigley, J. and Shiller, R. 2001. Comparing wealth effects: the stock market versus the
housing market, NBER Working Paper Series, No. 8606.

Coleman IV, M., LaCour-Little, M. and Vandell, K. 2008. Subprime lending and the housing
bubble: Tail wags dog?, Journal of Housing Economics, 17, 272 – 90.

Crowe, C., Dell’Ariccia, G., Igan, D. and Rabanal, P. 2011. How to deal with real estate booms:
Lessons from country experiences, IMF Working Paper, No. 11/91.

Davis, M.A., Lehnert, A. and Robert, F.M. 2008. The rent–price ratio for the aggregate stock of
owneroccupied housing, Review of Income and Wealth, 54, 279 – 84.

Dotsey, M. 1998. The predictive content of the interest rate term spread for future economic
growth, Federal Reserve Bank of Richmond Quarterly, 84, 31 – 51.

Dueker, M.J. 1997. Strengthening the case for the yield curve as a predictor of U.S Recessions,
Federal Reserve Bank of St. Louis Review, 79, March/April, 41 – 51.

Englund, P. and Ioannides, Y.M. 1997. House price dynamics: An international empirical
perspective, Journal of Housing Economics, 6, 119 – 36.

Estrella, A. and Hardouvelis, G.A. 1991. The term structure as a predictor of real economic
activity, Journal of Finance, 46, 555 – 76.

17
Estrella, A. and Mishkin, F.S. 1998. Predicting US recessions: Financial variables as leading
indicators, Review of Economics and Statistics, 80, 45 – 61.

Ghent, A.C. and Owyang, M.T. 2010. Is housing the business cycle? Evidence from U.S. cities,
Journal of Urban Economics, 67, 336 – 51.

Glindro, E.T., Subhanij, T., Szeto, J. and Zhu, H. 2011. Determinants of house prices in nine
Asia-Pacific Economies, International Journal of Central Banking, 3, 163 – 204.

Guidolin, M. and Timmermann, A. 2008. Asset allocation under multivariate regime switching,
Journal of Economic Dynamics and Control, 31, 3503 – 44.

Hall, S., Psaradakis, Z. and Sola, M. 1997. Switching error-correction models of house prices in
the United Kingdom, Economic Modelling, 14, 517 – 27.

Hamilton, J.D. 1989. A new approach of the economic analysis of nonstationary time series and
the business cycle, Econometrica, 57, 357 – 84.

Hamilton, J.D. 1994. Time series analysis. Princeton University Press, Princeton, NJ.

Hamilton, J.D. and Kim, D.H. 2000. A re-examination of the predictability of economic activity
using the yield spread, NBER Working Paper Series, No. 7954.

Haubrich, J.G. and Dombrosky, A.M. 1996. Predicting real growth using the yield curve, Federal
Reserve Bank of Cleveland Economic Review, 32, 26-35.

Himmelberg, C., Mayer, C. and Sinai, T. 2005. Assessing high house prices: Bubbles,
fundamentals and misperceptions, Journal of Economic Perspectives, 19, 67 – 92.

Holly, S. and Jones, N. 1997. House prices since the 1940s: Cointegration, demography and
asymmetries, Economic Modelling, 14, 549 – 65.

Hort, K. 1998. The determinants of urban house price fluctuations in Sweden: 1968 – 1994,
Journal of Housing Economics, 7, 93 – 120.

Iacoviello, M. and Minetti, R. 2003. Financial liberalization and the sensitivity of house prices to
monetary policy: Theory and evidence, Manchester School, 71, 20 – 34.

Kim, C.J. 1994. Dynamic linear models with markov-switching, Journal of Econometrics, 60, 1 – 22.

Lastrapes, W.D. 2002. The real price of housing and money supply shocks: Time series evidence
and theoretical simulations, Journal of Housing Economics, 11, 40 – 74.

Levin, E.J. and Wright, R.E. 1997. The impact of speculation on house prices in the United
Kingdom, Economic Modelling, 14, 567 – 85.

Ling, D. and Naranjo, A. 1999. The Integration of Commercial Real Estate Markets and the
Stock Market. Real Estate Economics, 27, 483 – 516.

McQuinn, K. and O’Reilly, G. 2008. Assessing the role of income and interest rates in
determining house prices, Economic Modelling, 25, 377 – 90.

18
Nneji, O., Brooks, C. and Ward, C. (2013a). Commercial real estate and equity market bubbles:
Are they contagious to REITs? Urban Studies, forthcoming.

Nneji, O., Brooks, C. and Ward, C. (2013b). Intrinsic and rational speculative bubbles in the US
housing market: 1960 – 2011, Journal of Real Estate Research, forthcoming.

Ortalo-Magne, F. and Rady, S. 2006. Housing market dynamics: On the contribution of income
shocks and credit contraints, Review of Economic Studies, 73, 459 – 485.

Ryden, T., Terasvirta, T. and Asbrink, S. 1998. Stylized facts of daily return series and hidden
markov model, Journal of Applied Econometrics, 13, 217 – 244.

Shiller, R. 2007. Understanding recent trends in house prices and home ownership, NBER
Working Paper Series, No. 13553.

Tsatsaronis, K., and Zhu, H. 2004. What drives housing price dynamics: Cross-country evidence,
BIS Quarterly Review, 65 – 78.

Whaley, R.E. 2000. The investor fear gauge, Journal of Portfolio Management, 26, 12 – 17.

Wheelock, D.C. 2006. What happens to banks when house prices fall? U.S. regional housing
busts of the 1980s and 1990s, Federal Reserve Bank of St. Louis Review, 413 – 28.

Xiao, Q. 2007. What drives Hong Kong’s residential property market – A Markov switching
present value model, Physica A, 383, 108 – 14.

19