Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Contents lists available at ScienceDirect
Cities
journal homepage: www.elsevier.com/locate/cities
The long- and short-run influences of housing prices on migration
Chien-Wen Peng
a,⁎
, I-Chun Tsai
b
a
Department of Real Estate and Built Environment, National Taipei University, Taipei, Taiwan
b
Department of Finance, National University of Kaohsiung, Kaohsiung, Taiwan
ARTICLE INFO
Keywords:
Housing price
Residential migration
Housing equity
Lock-in effect
ABSTRACT
A change in housing prices has a profound impact on households' housing equity and future moving decisions.
While most previous studies focus on discussing the mobility lock-in effect due to housing price depreciation,
revealing that there is a positive relationship between housing prices and migration, this study reexamines their
relationship by using the panel cointegration method and city-level panel data for Taiwan during the 1994–2016
period. The empirical results reveal that migration and housing prices are cointegrated, and the influence of
housing prices on migration is significantly positive in the long run. However, the influence of housing price
changes on migration is not as significant as expected in the short run. To further examine their short-run
relationships, we use quantile regression and the results show that the influence of housing price changes on
migration is significantly negative below the 0.5 quantile, but it turns out to be significantly positive in the 0.9
quantile. The influence of housing price changes on migration is not significant between the 0.5 and 0.8
quantiles. We conclude that the influence of housing price changes on migration might be asymmetric in the
short run.
1. Introduction
The changes in the population of a region are affected by both
natural and social factors. Natural factors include the difference be-
tween the number of newborns and deaths. Social factors comprise the
difference between the numbers of those moving in and those moving
out. Since the decision to move consists of many tangible (e.g., moving
expenses, relocation loss, brokerage fees, and tax) and intangible (e.g.,
the loss of current neighborhood networks, and re-adapting to a new
environment) costs, it is without doubt a very costly family decision.
Theoretically, all households will rationally evaluate the potential
benefits and costs before deciding to move. Only when the benefits are
greater than the costs will the decision to move be executed (Weinberg,
Friedman, & Mayo, 1981).
There are many factors that may affect a household's decision to
move, such as better housing quality, school quality, a better neigh-
borhood, increased accessibility, a change of housing tenure, and other
job-related reasons. When the moving distance is shorter, the reasons
for moving are more housing-related. (Ermisch & Washbrook, 2012;
Jones, Leishman, & Watkins, 2004;Peng, Wu, & Kung, 2009;Weinberg,
1979). When the moving distance is longer, the reasons for moving are
more job-related (Berger & Blomquist, 1992;Gabriel, Shack-Marquez, &
Wascher, 1992;Potepan, 1994;Zabel, 2012).
Households can improve their welfare by voluntarily deciding to
move. However, there are many factors that constrain households from
making such a decision, and the most important one is housing af-
fordability. A change in housing prices affects not only the cost of the
new home, but also the equity in the current home. Understanding how
the change in the housing price affects the household's moving decision
is an important task when making housing policy decisions, especially
in regions and countries with high home ownership rates.
A number of earlier studies focused on the influence of falling
housing prices on a household's moving decision. They argued that
falling housing prices will reduce a household's housing equity. In some
cases, the current market price may fall below the household's mort-
gage debt thereby giving rise to negative equity. A falling housing price
restricts the household's financial ability and reduces population mo-
bility. It results in the so-called equity lock-in effect (Blozea & Skak,
2016;Bricker & Bucks, 2016;Chan, 2001;Engelhardt, 2003;Ferreira,
Gyourko, & Tracy, 2010, 2012;Foote, 2016;Han, 2010;Modestino &
Dennet, 2013;Seko, Sumita, & Naoi, 2012;Sterk, 2015) Furthermore,
some studies have argued that a rise in housing prices increases a
household's housing equity, making it easier for households to move
(Disney, Gathergood, & Henley, 2010;Kiel, 1994). According to the
above-mentioned studies, the influence of the housing price on migra-
tion is positive.
https://doi.org/10.1016/j.cities.2019.05.011
Received 10 August 2018; Received in revised form 2 March 2019; Accepted 8 May 2019
⁎
Corresponding author at: 151, University Rd., San Shia District, New Taipei City, 23741, Taiwan.
E-mail addresses: cwpeng@mail.ntpu.edu.tw (C.-W. Peng), ictsai@nuk.edu.tw (I.-C. Tsai).
Cities 93 (2019) 253–262
Available online 04 June 2019
0264-2751/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
Although, most empirical studies confirm that the equity lock-in
effect arises from falling house prices, there are other studies that ob-
tain different results. They argue that the relationship between house-
holds moving and falling housing prices is not significant (Berger &
Blomquist, 1992;Schulhofer-Wohl, 2012;Valletta, 2013).
This study argues that the influence of housing prices on migration
might be asymmetric. As the price of housing rises, the households'
housing equity will increase, but their housing affordability might not
increase due to the relative growth of housing prices in different re-
gions. Furthermore, if households expect housing prices to continue to
increase in the near future, they might delay their moving decision and
cause moving to decrease. This means that the influence of housing
prices on the decision to move will be different during upturns and
downturns in the housing cycle and in the short run and long run.
The remainder of this paper is organized as follows. The next section
reviews the influences of the housing price on housing equity and mi-
gration. The third section consists of the research design and introduces
our hypotheses and the research methodology. The fourth section
presents the empirical results and discussion. Finally, we provide the
conclusion in the last section.
2. Literature review
Housing is the most important form of consumption and also in-
vestment to most households. A change in the price of housing will not
only influence a household's ability to afford a house but also the value
of the household's assets. Han (2010) examines how price risk affects
the demand for housing. He identifies two relevant channels: a financial
risk effect that reduces demand, and a hedging effect that increases
demand since current homes may act as a hedge against future housing
costs. For households with weak hedging incentives, the article finds
evidence of negative effects of price risk on the timing and size of home
purchases, but positive effects for households with strong hedging in-
centives.
As to the influence of changes in housing prices on population mi-
gration, most previous studies focused on whether falling housing
prices blocked households' moving decisions giving rise to the so-called
“lock-in effect”. From the empirical results of those studies, it can be
seen that most of them confirmed that falling housing prices did reduce
migration. Stein (1995) presents a simple model of trade in the housing
market. The crucial feature is that a minimum down-payment is re-
quired for the purchase of a new home. A fall in housing prices de-
creases the households' housing equity, so that the households are un-
able to afford the down-payment for a new home and thus their
decision to move is blocked. Chan (2001) argues that falling house
prices have caused numerous homeowners to suffer capital losses, and
those with little home equity may be prevented from moving because of
imperfections in the housing finance markets. Estimates show that if
house prices had not declined, average mobility would have been 24%
higher after 3 years, and would have been 33% higher after 4 years.
Among those with high initial loan-to-value ratios, the differences are
even greater.
Seko et al. (2012) investigate the effects of housing equity con-
straints and income shocks on own-to-own residential moves in Japan.
By looking at contemporaneous extended loan-to-value (ELTV) and
extended debt-to-income (EDTI) ratios under the recourse loan system,
they find that housing equity constraints and negative income shocks
significantly deter own-to-own residential moves for positive equity
households. Foote (2016) estimates the extent to which negative house
price changes lower mobility for some homeowners. He argues that
house price declines cause migration to decrease for homeowners that
have low equity, but that there is no effect for the most leveraged
homeowners. Differences in default costs across states do not appear to
affect the mobility of homeowners in negative equity. Housing lock-in
effects are observed to be larger during the most recent recession, af-
fecting both in-state and interstate migration.
Furthermore, Bricker and Bucks (2016) use panel data from the
2007–09 Survey of Consumer Finances to examine decisions by U.S.
households to move during the Great Recession and also look into the
role of negative home equity and economic shocks, such as job losses, in
these decisions. Their results show that many involuntary moves appear
to stem from a combination of negative home equity and adverse eco-
nomic shocks rather than from negative equity alone. The findings
suggest that, analogous to the double-trigger theory of default, the re-
lationship between negative equity and household mobility varies with
the exposure of households to adverse shocks.
Some studies analyze the influence of rising housing prices on
household mobility. Kiel (1994) empirically tests how both prior and
future appreciation affect households' moving decisions by using
American Housing Survey data and a nonparametric estimation tech-
nique. He finds that homeowners over the age of 40 with more than five
years in their unit were more likely to move if their unit experienced
higher-than-average future appreciation and if their unit had experi-
enced higher-than-average appreciation in the past. The latter result
indicates that repeat-sales indices may be upwardly biased; the former
indicates that households may not correctly estimate future apprecia-
tion. Disney et al. (2010) examine the impact of unanticipated housing
capital gains on consumption behavior using data from the British
Household Panel Survey and county-level house price data. They find a
marginal propensity to consume out of unanticipated shocks to housing
wealth of 0.01 and there is little evidence of heterogeneity in the re-
sponses of young and old homeowners, but there are differences
Table 1
Definitions of variables and data sources.
Variables Definition Data source Exp. sign
Dependent variable
Migration Rate (MR) The sum of population moving to other cities, other districts and also within the same
districts divided by the total population in that city
Statistics Office, Ministry of Interior,
Taiwan, R.O.C.
Gross Migration Rate (GMR) The sum of population moving in and out of the same districts, different districts and
also different cities divided by the total population in that city
Same as above
Intra-City Migration Rate of
(ICMR)
The sum of population moving within the same and to different districts of the same city
divided by the total population in that city
Same as above
Intra-District Migration Rate
(IDMR)
The sum of population moving within the same districts of the same city divided by the
total population in that city
Same as above
Independent variables
Proportion of Married Couples
(PMC)
The proportion of married people to population aged 15 and above Same as above +
Elderly Population Ratio (EPR) The proportion of people aged 65 and above to total population Same as above −
Household Income (HI) Average household income in each city and county Same as above +
Housing Stock (HS) Total number of housing units in each city and county Same as above −
Housing Price (HP) Average housing price in each city and county Calculated by the authors +/−
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
254
between owners and renters. They also find asymmetric behavior be-
tween house price rises and falls, and a disproportionate impact on
saving if the household had negative housing equity at the start of the
period.
Ermisch and Washbrook (2012) present a model in which housing
equity can influence mobility. They estimate parameters that gauge the
impact of housing equity, local house prices and other variables asso-
ciated with household structure and change on residential moves within
the UK. Their main finding is that an increase in a household's housing
equity encourages residential mobility substantially, and a decline
discourages it. Hendershott, Lee, and Shilling (2015) analyze the im-
pact of dramatic changes in housing prices on housing turnover. They
find that there is an equity lock-in effect for homeowners with a high
loan-to-value ratio when housing prices fall, and also an interest rate
lock-in effect for the homeowners with adjustable rate mortgages when
interest rates move upward. Both the equity lock-in and interest rate
lock-in effect reduce turnover and household mobility.
However, there are still some studies that find the influence of
housing prices on migration to be insignificant. For example, Berger
and Blomquist (1992) consider the individual's decision to move and
choice of destination. They find that wages and moving costs are most
important when choosing whether or not to move. Quality of life,
wages, and housing prices matter in the choice of destination.
Valletta (2013) analyzes whether the recent decline in internal
migration in the United States may have been caused in part by falling
house prices through the “lock-in”effects of the financial constraints
faced by households whose housing debt exceeds the market value of
their homes. He examines the relationship between such “lock-in”ef-
fects and the elevated levels and persistence of unemployment during
the recent recession and its aftermath, using data for the years 2008–11,
with a special focus on differences in unemployment duration between
homeowners and renters across geographic areas differentiated by the
severity of the decline in home prices. However, he does not find sys-
tematic evidence to support the house “lock-in”effect hypothesis.
Peng et al. (2009) argue that a higher intra-regional mobility rate
may indicate the fact that households in the region can be more ef-
fective in raising the quality of life and meet their needs by moving to
another house in the same region. By employing panel data covering 22
localities in Taiwan over the 1982–2007 period, they find that the intra-
regional mobility rate is negatively associated with the homeownership
rate, but positively associated with the marriage rate, floor area occu-
pied, vacancy rate, and the ratio of the housing price to income. Both
the homeownership rate and marriage rate are among the key de-
terminants.
Since the empirical results of previous studies related to the influ-
ence of housing prices on migration are still ambiguous, it might be the
case that the influence of housing prices on migration is asymmetric. It
is important to further investigate the influence of housing prices on
migration during upturns and downturns in the housing cycle and also
the short run and long run.
3. Research design
The quality and quantity of public services differ quite significantly
among regions, cities and administrative districts. A change in housing
prices will influence both homeowners and renters and cause them to
engage in interregional and intraregional population migration.
As for the homeowners, a rise in housing prices will increase their
housing equity and improve their ability financially to relocate. So, the
influence of housing prices on migration is positive. However, whether
the homeowners can move depends not just on the changes in the prices
of their homes, but also on the changes in the housing prices in the
region to which they intend to move. If the relative growth of housing
prices in their current region over other regions is lower, they may not
be able to move and the outmigration rate will be reduced. Conversely,
if the relative growth of the housing prices in their region is greater
Table 2
Basic statistics.
Dependent variables MR GMR ICMR IDMR
Mean 8.6614 17.2192 5.3234 2.9577
Std. deviation 2.2992 4.7349 1.5756 0.7800
Minimum 4.6200 8.7800 1.8500 0.9500
Maximum 16.0500 33.3100 11.4900 5.5000
Independent Variables HP HI PMC EPR HS
Mean 9.9175 658,759.5 0.5595 0.0871 290,667.7
Std. deviation 6.8513 268,631.1 0.0373 0.0309 279,349.3
Minimum 1.7800 155,795.2 0.4490 0.0290 49,652
Maximum 61.2268 1,320,834 0.6908 0.1790 1,596,779
Notes: MR is migration rate. GMR is gross migration rate. ICMR is intra-city migration rate. IDMR is intra-district migration rate. HP is housing price. HI is household income. PMC is proportion of
married couples. EPR is elderly population ratio. HS is housing stock.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
255
Fig. 1. Migration rates of the different cities.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
256
than that in the other regions, the households will be more capable of
moving. This will increase the outmigration rate and also reduce the
immigration rate. Furthermore, expectations regarding future housing
prices will also influence households' moving decisions. If the ex-
pectation is that housing prices will grow faster in the current region
than in the region to which the households intend to move, then they
will choose not to move in the short run and this will reduce the out-
migration rate.
As to the renters, if their incomes are stable, a rise in housing prices
will weaken their ability to purchase a home. The current renters can
just keep on renting or move to other regions with lower housing prices.
The net effect of housing price changes on their moving decisions will,
however, be unclear.
The reasons for households moving can be simply divided into re-
sidential and nonresidential migration. According to the moving dis-
tance and whether they are moving across different administrative
districts, we divide the residential migration into three types:
1. Migration within the same district, in the same city: It is the shortest
distance for residential migration, and the differences in terms of
neighborhoods, housing prices and rents are relatively small.
2. Migration across different districts within the same city: The moving
distance is greater than that of moving within the same district, and
the differences in neighborhoods, housing prices and rents are also
greater.
3. Migration across different cities: The moving distance is the longest
in terms of residential migration, and the differences in neighbor-
hoods, housing prices and rents are also the largest.
3.1. Models
This study uses the panel cointegration tests developed by Pedroni
(1999, 2004) to examine the relationships between the migration rate
and housing price, and other migration-related variables. The basic
model is as follows:
=+ + +…+ +…+ +
…
YαλtβX βX βX ε.
it i i iit mi mi t Mi Mi t i t,11, , , ,
(1)
In Eq. (1),Y
i,t
is the dependent variable, the X
mi,t
are the in-
dependent variables, β
1i
,β
2i
,….β
Mi
are coefficients of X
mi,t
,α
i
is the
intercept, and λ
i
tis the long-term trend. Furthermore, α
i
and β
1i
,β
2i
,….
β
Mi
also denote the cointegration vector, which captures the linear re-
lationship between the dependent variable and independent variables,
and the ε
i,t
are the error terms, which represent the deviation between
the dependent variables and the long-term equilibrium. If the error
term is stationary, then it reveals a long-term equilibrium relationship
between Y
i,t
and the other independent variables. For completely
analyzing the determinants of the migration rate, the three types of
residential migration rates are used to be the dependent variable in Eq.
(1). Housing price is the main independent variable, and based on the
findings of the previous studies, we also use household income, elderly
people ratio, proportion of married couples, and housing stock as the
explanatory variables to estimate Eq. (1).
The panel cointegration tests of Pedroni (1999, 2004) in Eq. (1) can
also be used to test heterogeneous panels because they consider the
heterogeneous slope coefficients, fixed effects and the trends in in-
dividual sectors. The heterogeneous slope coefficients in different sec-
tors (β
1i
,β
2i
,….β
Mi
), the intercept (α
i
) and trend (λ
i,t
) can be hetero-
geneous; thus, the requirements are more flexible than those of the
homogeneous panel model. We can also examine whether the error
term is stationary by computing Eq. (2), as follows:
=+
−
εθε ν
it i it i
t
,,1, (2)
The null hypothesis is θ
i
= 1; if it is rejected, then there is a coin-
tegration relationship. Two statistics are used to measure the coin-
tegration relationship; both were developed by Pedroni (1999). The
first one is panel cointegration. The rejection of the null hypothesis
Table 3
Unit root test.
Dependent variables MR GMR ICMR IDRT
In level 20.5653
(0.9990)
23.1513
(0.9959)
56.7438
(0.0942)
52.6710
(0.1737)
In difference 1165.7200
(0.0000)
1062.8400
(0.0000)
1225.6500
(0.0000)
823.5620
(0.0000)
Independent variables HP HI PMC EPR HS
In level 24.7597
(0.9915)
40.1904
(0.6356)
13.3115
(1.0000)
8.0331
(1.0000)
28.8627
(0.9620)
In difference 234.7310
(0.0000)
509.9110
(0.0000)
379.3460
(0.0000)
141.6470
(0.0000)
337.5780
(0.0000)
Notes: MR is migration rate. GMR is gross migration rate. ICMR is intra-city migration rate. IDMR is intra-district migration rate. HP is housing price. HI is household income. PMC is proportion of
married couples. EPR is elderly population ratio. HS is housing stock. p-value are in parentheses.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
257
indicates that a cointegration relationship exists in all sectors. The
second one is group mean panel cointegration. If there is a cointegra-
tion relationship in any sector then the statistic will significantly reject
the null hypothesis.
3.2. Variables
We refer to previous studies to choose the housing price (HP),
household income (HI), elderly people ratio (EPR), proportion of mar-
ried couples (PMC), and housing stock (HS) as the explanatory variables
for our migration model.
The incentive for elderly to move is low, so an increase in the el-
derly people ratio will decrease the migration rate. As the housing stock
increases, the problem of a housing shortage will decrease, so the mi-
gration rate will be lower. As the household income increases, the
ability to move to a better environment will also increase, so its influ-
ence on the migration rate is positive. As the proportion of married
couples increases, the demand for location and space adjustment will
increase, and so its influence on migration will also be positive. The
definitions and sources of the variables are described in Table 1.
4. Empirical results
4.1. Basic statistics
This study employs the panel data of 19 cities and counties in
Taiwan for the period from 1994 to 2016, which include all the cross-
sectional and time series information. The basis statistics are shown in
Table 2.Fig. 1 shows the trend of the migration for different cities.
Although all migration rates of different cities declined in the sample
period, there were still different time patterns in these migration rates.
4.2. Long-run equilibrium
First, we use the unit root test to examine the characteristics of all
dependent and independent variables. According to Table 3, none of the
variables reject the null hypothesis of having a unit root in levels, which
means that they are non-stationary. However, all the variables sig-
nificantly reject the null hypothesis in terms of the first differences. We
conclude that all migration-related variables are I(1), so that we can
examine their long-run relationships by using the cointegration test.
The results of the cointegration test between migration and other
variables are presented in Table 4. According to the results of the panel
cointegration test and group mean panel cointegration test, the null
hypothesis is significantly rejected. We conclude that all migration
types are cointegrated with our selected variables in all cities.
This study measures the cointegration vectors by using the Fully
Modified OLS (FMOLS) model. According to the results shown in
Table 5, it can be seen that all variables significantly influence the
migration rate in the long run. The fitness of the model is quite good
and the signs of the coefficients are all as expected. The impacts of EPR
and HS on the migration rate are significantly negative. It means that
when the elderly population ratio increase in a city, the moving po-
pulation will decrease. Likewise, when the total number of housing
units increases in a city, the households are easier to find a suitable
home, and then decrease their moving behavior. Conversely, the in-
fluences of HP, HI, and PMC on the migration rate are positive. It means
that an increase of household income will make households more af-
fordable to move. An increase of proportion of married couples will
change households' housing demand in different family life cycles, and
then increase households' moving. An increase of housing price will also
increase households' burden to purchase a home, however, since most
of the households are homeowners, the increase of housing equity may
compensate this disadvantage.
The influence of the housing price on the migration rate is positive
in the long run, which is in accordance with the previous migration
“lock-in”effect studies. As the housing price falls, the migration rate
will become lower. Furthermore, migration rate in a housing expensive
region will be relatively higher than in a housing affordable region.
High housing prices are usually caused by both strong housing demand
Table 4
Cointegration test.
Statistics p-value Weighted statistics p-Value
(a)MR,HP,HI,PMC,EPR,HS
Alternative hypothesis: common AR coeffs. (within-dimension)
Panel PP-Statistic −9.6702 0.0000 −13.6173 0.0000
Panel ADF-Statistic −8.1097 0.0000 −8.9017 0.0000
Alternative hypothesis: individual AR coeffs. (between-dimension)
Group PP-Statistic −23.1232 0.0000
Group ADF-Statistic −12.3201 0.0000
(b)GMR,HP,HI,PMC,EPR,HS
Alternative hypothesis: common AR coeffs. (within-dimension)
Panel PP-Statistic −8.9077 0.0000 −15.3047 0.0000
Panel ADF-Statistic −7.7440 0.0000 −8.8569 0.0000
Alternative hypothesis: individual AR coeffs. (between-dimension)
Group PP-Statistic −21.5864 0.0000
Group ADF-Statistic −12.1315 0.0000
(c)ICMR,HP,HI,PMC,EPR,HS
Alternative hypothesis: common AR coeffs. (within-dimension)
Panel PP-Statistic −9.9499 0.0000 −19.1658 0.0000
Panel ADF-Statistic −8.2803 0.0000 −11.1920 0.0000
Alternative hypothesis: individual AR coeffs. (between-dimension)
Group PP-Statistic −23.1132 0.0000
Group ADF-Statistic −12.2931 0.0000
(d)IDMR,HP,HI,PMC,EPR,HS
Alternative hypothesis: common AR coeffs. (within-dimension)
Panel PP-Statistic −10.6905 0.0000 −11.5119 0.0000
Panel ADF-Statistic −8.7148 0.0000 −9.1734 0.0000
Alternative hypothesis: individual AR coeffs. (between-dimension)
Group PP-Statistic −17.1488 0.0000
Group ADF-Statistic −10.7999 0.0000
Notes: MR is migration rate. GMR is gross migration rate. ICMR is intra-city migration rate. IDMR is intra-district migration rate. HP is housing price. HI is household
income. PMC is proportion of married couples. EPR is elderly population ratio. HS is housing stock.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
258
Table 5
Cointegration vector.
Variable Coefficient Std. error t-Value p-Value
Dependent variable: MR
HP 0.0579 0.0279 2.0783 0.0382
HI 0.0000 0.0000 3.6048 0.0003
PMC 18.1901 1.5572 11.681 0.0000
EPR −44.4619 4.8511 −9.1654 0.0000
HS −0.0000 0.0000 −2.5953 0.0097
R-squared 0.3537
Adjusted R-squared 0.3483
S.E. of regression 1.8561
Long-run variance 8.3970
Mean dependent var 8.6614
S.D. dependent var 2.2992
Sum of squared resid 1663.9370
Dependent variable: GMR
HP 0.0912 0.0546 1.6703 0.0955
HI 0.0000 0.0000 4.6389 0.0000
PMC 35.1661 3.0499 11.5304 0.0000
EPR −96.1732 9.5007 −10.1227 0.0000
HS −0.0000 0.0000 −2.5342 0.0116
R-squared 0.4128
Adjusted R-squared 0.4080
S.E. of regression 3.6432
Long-run variance 32.2082
Mean dependent var 17.2192
S.D. dependent var 4.7349
Sum of squared resid 6410.7880
Dependent variable: ICMR
HP −0.0095 0.0200 −0.4754 0.6347
HI 0.0000 0.0000 5.7483 0.0000
PMC 8.7008 1.1152 7.8020 0.0000
EPR −25.7078 3.4740 −7.4000 0.0000
HS −0.0000 0.0000 −1.3139 0.1895
R-squared 0.2755
Adjusted R-squared 0.2695
S.E. of regression 1.3466
Long-run variance 4.3064
Mean dependent var 5.3234
S.D. dependent var 1.5756
Sum of squared resid 875.8856
Dependent variable: IDMR
HP 0.0064 0.0095 0.6733 0.5011
HI 0.0000 0.0000 5.0611 0.0000
PMC 5.4243 0.5313 10.2100 0.0000
EPR −14.0051 1.6550 −8.4622 0.0000
HS −0.0000 0.0000 −1.7365 0.0831
R-squared 0.3210
Adjusted R-squared 0.3154
S.E. of regression 0.6454
Long-run variance 0.9774
Mean dependent var. 2.9577
S.D. dependent var. 0.7800
Sum of squared resid. 201.2017
Notes: MR is migration rate. GMR is gross migration rate. ICMR is intra-
city migration rate. IDMR is intra-district migration rate. HP is housing
price. HI is household income. PMC is proportion of married couples.
EPR is elderly population ratio. HS is housing stock.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
259
and inelastic land supply. It is quite difficult for most households to get
their ideal home by just one or few moving decisions under expensive
housing prices. They need to adjust their housing demand through
many moving behaviors, so that the migration rate will be higher.
Conversely, the households have much less moves in a relatively
housing affordable region, so the migration rate will be lower. We have
added the discussion of this in the revised manuscript.
The effects of the housing price on GMR, ICMR and IDMR are not
significant, which contradicts the results in Table 4. This result might be
due to the close relationship between household income and the
housing price, because some important information on the housing
price has already been reflected by household income. This result can
also reflect the advantage of the FMOLS model in terms of solving the
problem of multicollinearity among the independent variables.
4.3. Short-run dynamics
Table 5 shows that the influences of the housing price and house-
hold income on migration differ among migration types. Since the
number of significant variables is more for MR than for the others, we
will only focus on the influence of the housing price on MR in the
following analysis.
Table 6 shows the results of the error correction model. The error
correction term is obtained from the FMOLS model shown in Table 5.
The coefficient of the error correction term is significantly negative. It
reveals that when the migration rate is higher than the long-run equi-
librium that is estimated by the other variables, it will adjust downward
to the long-run equilibrium in the next period. Furthermore, the in-
fluence of a one-period lead of ΔEPR on ΔMR is significantly negative,
but the influence of other variables on the migration rate is not sig-
nificant.
We find that the effects of ΔEPR on migration are significant in both
the long and short run. However, the influence of the housing price
changes on migration is not significant, which is not in accordance with
previous studies. Is it possible that the influence of the housing price on
migration is asymmetric?
Similar to recently developed panel data unit root and causality
tests, there is an analogous test for the Granger causality in the panel
data with a short time-series dimension. This test is described in Hurlin
(2005) and applied in Hurlin and Venet (2008).
Let γ
i
be the endogenous variable denoting house prices (household
income) and η
i
be the regression variable denoting household income
(house prices). Consider the following linear model:
∑∑
=+ + +…
=
−
=
−
γμ ϕγ ψη δ
Δ
ΔΔ
it i
l
L
i
lit l
l
L
i
lit l it
,
1
()
,
1
()
,,(3)
where δ
i,t
are normally i.i.d. with zero mean and finite heterogeneous
variances, and δ
i
=(δ
i,t
,…,δ
i,T
)
′
are independently distributed across
groups. The null hypothesis assumes that ηdoes not help predict γfor
any of the Nindividual units in the panel. It is referred to as
Homogeneous Non-causality and can be formally stated as:
==∀=…Hψ N0, 1, ,
.
ii0
We examine the causal relationship between housing price change
and migration rate change by using Pairwise Dumitrescu Hurlin Panel
Table 6
Error correction model.
Variable Coefficient Std. error t-Value p-Value
Dependent variable: ΔMR
ECT
t−1
−0.1578 0.0234 −6.7562 0.0000
ΔHP
t−1
−0.0121 0.0376 −0.3220 0.7476
ΔHI
t−1
0.0000 0.0000 −0.8356 0.4038
ΔPMC
t−1
−7.3602 6.0488 −1.2168 0.2243
ΔEPR
t−1
−48.3981 16.4000 −2.9511 0.0033
ΔHS
t−1
0.0000 0.0000 −0.0387 0.9691
R-squared 0.0839
Adjusted R-squared 0.0739
S.E. of regression 0.9306
Log likelihood −624.6803
Mean dependent var. −0.1698
S.D. dependent var. 0.9670
Sum of squared resid 398.3535
Notes: ECT is error correction term. MR is migration rate. HP is housing
price. HI is household income. PMC is proportion of married couples.
EPR is elderly population ratio. HS is housing stock.
Table 7
Pairwise Dumitrescu Hurlin Panel Causality.
Null Hypothesis W-Statistic Zbar - statistic p-Value
ΔMR does not homogeneously cause
ΔHP
t−1
0.9020 −0.6120 0.5405
ΔHP
t−1
does not homogeneously cause
ΔMR
0.4578 −1.7778 0.0754
Notes: MR is migration rate. HP is housing price.
Table 8
Quantile regression model.
Dependent variable: ΔMR
Ind. Var. Quantile Coefficients Std. error t-Value p-Value
ECT
t−1
0.1 −0.1160 0.0145 −7.9891 0.0000
0.2 −0.1074 0.0193 −5.5732 0.0000
0.3 −0.0999 0.0222 −4.4934 0.0000
0.4 −0.0856 0.0235 −3.6450 0.0003
0.5 −0.0758 0.0234 −3.2365 0.0013
0.6 −0.0630 0.0228 −2.7618 0.0060
0.7 −0.0554 0.0213 −2.5981 0.0097
0.8 −0.0521 0.0184 −2.8398 0.0047
0.9 −0.0404 0.0145 −2.7845 0.0056
ΔHP
t−1
0.1 −0.0932 0.0160 −5.8379 0.0000
0.2 −0.0841 0.0207 −4.0604 0.0001
0.3 −0.0689 0.0241 −2.8640 0.0044
0.4 −0.0545 0.0266 −2.0490 0.0410
0.5 −0.0325 0.0297 −1.0923 0.2753
0.6 −0.0055 0.0280 −0.1958 0.8449
0.7 0.0037 0.0255 0.1460 0.8840
0.8 0.0231 0.0230 1.0042 0.3158
0.9 0.0614 0.0233 2.6388 0.0086
Notes: MR is migration rate. ECT is error correction term. HP is housing price.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
260
Causality test. Table 7 shows that the change of migration rate could
not granger cause the variations of housing price, so our discussion only
focus on the influence of housing price change on migration rate
change.
Furthermore, this study measures the long- and short-run influences
of the housing price on migration via quantile regression. Table 8 shows
that the migration rate is significantly influenced by the one-period lead
error term. When the error terms are too high, the migration rate will
adjust downward. The extent of the adjustment is more significant in
the lower quantiles. The influence of the housing price on migration is
positive in the long run.
However, the influence of the housing price changes on migration
might be asymmetric in the short run. When the migration rate changes
are lower (for example lower than the 0.5 quantile), the one period
leading the housing price changes significantly negatively influences
the migration rate changes. It means that high housing prices result in a
lower migration rate. It reveals that not only a falling but also a rising
housing price might lead to a migration lock-in effect. Moreover, the
influence of the housing price changes on migration becomes positive
when the migration rate is high (in the 0.9 quantile), which is in ac-
cordance with the long-run effect. The influence of the housing price
changes on migration is not significant from the 0.5 to 0.8 quantiles.
5. Conclusion
A change in the housing price has a great impact on households'
housing equity and future moving decisions. While most previous stu-
dies focus on discussing the mobility lock-in effect due to housing price
depreciation, this study reexamines their relationship by using the
panel cointegration method and city level panel data covering the
1994–2016 period in Taiwan. The empirical results reveal that migra-
tion and the housing price are cointegrated, and the influence of the
housing price on migration is significantly positive in the long run.
However, the influence of the housing price on migration is not as
significant as expected in the short run. We used quantile regression to
further examine their short-run relationships, and the results showed
that the influence of the housing price on migration is significantly
negative below the 0.5 quantile, but it turns to be significantly positive
in the 0.9 quantile. The influence of the housing price on migration is
not significant from the 0.5 to the 0.8 quantiles.
We conclude that the influence of the housing price on migration
might be asymmetric. When the housing price rises in one specific re-
gion and increases the housing equity of most households in that re-
gion, the housing prices in other regions might also increase due to the
ripple effect. For most households with only one housing unit, the
ability to move to other regions might not increase proportionately.
Furthermore, if the households expect housing prices to continue to
increase in the near future, they might continue to hold on to their
home instead of selling it and realizing the capital gain. Under such a
situation, the relationship between housing prices and migration might
be negative. One other possible outcome is that the housing transaction
volume will increase as the housing price goes up, but most buyers in
such cases will be investors instead of consumers. Rising housing prices
do not necessarily that household mobility will be triggered and
housing quality increased.
There are some suggestions for further research. Firstly, the ad-
vantage of panel data analysis is containing both cross section and time
series characteristics of the data which allowing us to examine the long
and short run relationships of the selected variables. However, there are
also some limitations to include all significant variables which used in
the micro data migration analysis because of the data availability.
Secondly, the key motivations behind households' migration decisions
are complicated. Using the city-level panel data can just measure the
aggregate net effect of housing price on migration among households,
but not able to distinguish heterogeneous behaviors and various me-
chanisms driving housing demand by renters, single-house owners and
multi-houses owners (investors). Thirdly, there is a theoretically strong
relationship between rental housing market and owner-occupied
housing market. However, the city-level home ownership rates are
around 80 to 90% and rent is relatively stable over time in Taiwan.
Households prefer owning versus renting due primarily to the low user
cost of owner-occupied housing, which is due in part to house price
inflation (Bourassa & Peng, 2011). We do not measure the effects of
rental housing market on migration, and this can be tested in further
research. Finally, we explain the asymmetric relationship between
housing price change and variation of migration rate in the short run by
using lock-in effect and ripple effect. These arguments are not directly
support by the empirical results and still need further tests to verify.
Acknowledgment
We thank the Ministry of Science and Technology, Taiwan for
funding support.
Funding
This study was supported by the Ministry of Science and
Technology, Taiwan (MOST106-2410-H-305-058-MY2).
References
Berger, M. C., & Blomquist, G. C. (1992). Mobility and destination in migration decisions:
The roles of earnings, quality of life, and housing prices. Journal of Housing Economics,
2,37–59.
Blozea, G., & Skak, M. (2016). Housing equity, residential mobility and commuting.
Journal of Urban Economics, 96, 156–165.
Bourassa, S. C., & Peng, C. W. (2011). Why is Taiwan's homeownership rates so high?
Urban Studies, 48(13), 2887–2904.
Bricker, J., & Bucks, B. (2016). Negative home equity, economic insecurity, and house-
hold mobility over the great recession. Journal of Urban Economics, 91, 1~12.
Chan, S. (2001) Spatial lock-in: Do falling house prices constrain residential mobility?
Journal of Urban Economics, 49(3), 567~86.
Disney, R., Gathergood, J., & Henley, A. (2010). House price shocks, negative equity, and
household consumption in the United Kingdom. Journal of the European Economic
Association, 8(6), 1179–1207.
Engelhardt, G. V. (2003). Nominal loss aversion, housing equity constraints, and house-
hold mobility: Evidence from the United States. Journal of Urban Economics, 53(1),
171–195.
Ermisch, J., & Washbrook, E. (2012). Residential mobility: Wealth, demographic and
housing market effects. Scottish Journal of Political Economy, 59(5), 483–499.
Ferreira, F., Gyourko, J., & Tracy, J. (2010). Housing busts and household mobility.
Journal of Urban Economics, 68(1), 34–45.
Ferreira, F., Gyourko, J., & Tracy, J. (2012). Housing busts and household mobility: An
update. FRBNY Economic Review, November,1–15.
Foote, A. (2016). The effects of negative house Price changes on migration: Evidence
across U.S. housing downturns. Regional Science and Urban Economics, 60, 292–299.
Gabriel, S. A., Shack-Marquez, J., & Wascher, W. L. (1992). Regional house-price dis-
persion and interregional migration. Journal of Housing Economics, 2, 235–256.
Han, L. (2010). The effects of price risk on housing demand: Empirical evidence from U.S.
markets. Review of Financial Studies, 23(11), 3889–3928.
Hendershott, P. H., Lee, J. M., & Shilling, J. D. (2015). The 2005–2011 housing boom and
bust: Impacts on housing turnover and implications for the recovery. Journal of Real
Estate Research, 37(4), 471–498.
Hurlin, C. (2005). Un Test Simple de L'hypothèse de Non-Causalité Dans un Modèle de
Panel Hétérogène. Revue Économique, 56(3), 799–809.
Hurlin, C., & Venet, B. (2008). Financial development and growth: A re-examination
using a panel granger causality test. HALSHS working paper No. 00319995. Retrieved
from https://hal.archives-ouvertes.fr/halshs-00319995/.
Jones, C., Leishman, C., & Watkins, C. (2004). Intra-urban migration and housing sub-
markets. Housing Studies, 19(2), 269–283.
Kiel, K. A. (1994). The impact of house price appreciation on household mobility. Journal
of Housing Economics, 3(2), 92–108.
Modestino, A. S., & Dennet, J. (2013). Are American homeowners locked into their
houses? The impact of housing market conditions on state-to-state migration.
Regional Science and Urban Economics, 43, 322–337.
Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with
multiple regressors (Special Issue). Oxford Bulletin of Economics and Statistics, 61,
653–670.
Pedroni, P. (2004). Panel cointegration: Asymptotic and finite sample properties of
pooled time series tests with an application to the PPP hypothesis. Econometric
Theory, 20, 597–625.
Peng, C.-W., Wu, W. C., & Kung, S.-Y. (2009). An analysis of determinants of residential
migration. Journal of Population Studies, 39,85–118 (in Chinese with English ab-
stract).
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
261
Potepan, M. J. (1994). Intermetropolitan migration and housing prices: Simultaneously
determined. Journal of Housing Economics, 3,77–91.
Schulhofer-Wohl, S. (2012). Negative equity does not reduce homeowners' mobility.
Federal Reserve Bank of Minneapolis Quarterly Review, 35(1), 2–14.
Seko, M., Sumita, K., & Naoi, M. (2012). Residential mobility decisions in Japan: Effects
of housing equity constraints and income shocks under the recourse loan system.
Journal of Real Estate Finance and Economics, 45(1), 63–87.
Stein, J. C. (1995). Prices and trading volume in the housing market: A model with down-
payment effects. Quarterly Journal of Economics, 110(2), 379–406.
Sterk, V. (2015). Home equity, mobility, and macroeconomic fluctuations. Journal of
Monetary Economics, 74(3), 16–32.
Valletta, R. G. (2013). House lock and structural unemployment. Labour Economics, 25,
86–97.
Weinberg, D. H. (1979). The determinants of intra-urban household mobility. Regional
Science and Urban Economics, 9(2–3), 219–246.
Weinberg, D. H., Friedman, J., & Mayo, S. K. (1981). Intraurban residential mobility: The
role of transactions costs, market imperfections, and household disequilibrium.
Journal of Urban Economics, 9(3), 332–348.
Zabel, J. E. (2012). Migration, housing market, and labor market responses to employ-
ment shocks. Journal of Urban Economics, 72(2–3), 267–284.
C.-W. Peng and I.-C. Tsai Cities 93 (2019) 253–262
262
