Determinants of Housing Demand in Urban Areas of Pakistan: Evidence from the PSLM.
Ahmed, Ayaz ; Iqbal, Nasir ; Siddiqui, Rehana 等
Determinants of Housing Demand in Urban Areas of Pakistan: Evidence from the PSLM.
The study attempts to investigate the determinants of housing
demand in urban areas of Pakistan. The empirical analysis is carried out
using the Pakistan Social and Living Standard Measurement (PSLM) survey
2004-05 and 2010-11. The hedonic price model is used for the estimation
of house prices. In order to control the selectivity bias between the
tenure choice and the quantity of housing services demanded,
Heckman's two-step selection procedure is used. The empirical
analysis shows that housing price and income (temporary and permanent)
play an important role in the determination of the housing units'
demand. An increase in houses' prices causes decrease in demand for
the housing units while the housing demand increases when the permanent
income increases. On the face of change or increase in the transitory
income, the demand for housing units remains static, since people do not
desire to make long-term decisions based on volatile income. To manage
rising housing demand, government should focus on developing effective
and enforced price control mechanisms.
Keywords: Urban Housing Demand, PSLM, Pakistan
1. INTRODUCTION
Housing is a basic human need and millions of people struggle to
have a roof over their heads. In the face of unprecedented urbanisation
and population growth many cities have accrued huge housing shortages,
especially in developing and emerging economies. Estimates show that in
2010 around 980 million urban households lacked decent housing, as will
another 600 million between 2010 and 2030. One billion new homes will be
required worldwide by 2025, costing an estimated amount of $650 billion
per year [UN-Habitat (2016)]. Although hundreds of new housing colonies
have been established, the problem of finding a suitable accommodation
in big cities persists. In 2014, more than 30 percent of urban
population resided in slums in developing countries. Since every
household is not able to build a house for itself, there is always a
demand for rental houses. Housing, therefore, as a basic need became a
challenging outlay of rapid urbanisation in most of the developing
countries [UN-Habitat (2016)].
Housing demand is simply a housing need, which is backed up by the
ability and willingness to pay. It depends on the different forms of
behaviour of individuals, that how various households spend their
limited resources, to fulfil their needs of housing units as well as
their need for goods and services. The need of urban housing is affected
by a number of factors, such as: rural to urban migration, increasing
population, low investment in housing development, low purchasing power
of household, poor urban infrastructure and geography [Fontenla,
Gonzalez, and Navarro (2009); Oktay, Karaaslan, Alkan, and Kemal Celik
(2014); Saiz (2010)].
Pakistan is a developing country that accommodates the world's
sixth largest population. The housing situation has remained under
pressure in Pakistan. Pakistan has been confronting housing issues in
both qualitative and quantitative terms. Pakistan is faced with a severe
shortage of housing, particularly for low and middle-income groups.
Estimates disclosed that there is a shortage of about 7.5 million
housing units [SBP (2013)]. The gap between supply and demand is
increasing by more than 0.35 million. This issue is more critical in
urban regions, where accessibility of sufficient residences at
affordable rents is getting scarcer by the day. Population growth,
rising urbanisation and economic development have created huge housing
backlog, especially in big cities. This has not only increased the need
for new housing units, but has also created a huge burden on the
existing housing units. The existing work on housing in Pakistan by
Pasha and Ghaus (1990), Lodhi and Pasha (1991), Ahmed (1994) and Pasha
and Butt (1996), represents the first few attempts at determining
factors that affect housing demand. Very few studies have ventured into
determining the housing demand across income groups [Shefer (1990) and
Tiwari and Parikh (1998)].
Given this background, the prime objective of this study is to
analyse the housing demand in urban areas of Pakistan. This study
attempts to determine empirically, how the factors such as wealth,
income and house prices influence the ability to own a house differently
for low, middle and high-income groups. This study compares the
influence of socio-economic factors on the housing demand for two
different time periods i.e. 2004-05 and 2010-11. In 2004-05, the housing
market boomed while in 2010-11 the housing market was faced with
recession.
This study makes a significant contribution to literature in
various contexts. First, the study identifies all the possible factors,
affecting the housing demand at national, provincial and income
groups' levels. Second, this study compares the influence of
socioeconomic factors on housing demand for two different time periods
i.e. 2004-05 and 2010-11. In 2004-05, the housing market was at its boom
while in 2010-11 housing market was faced with recession. Both periods,
therefore, have different implications for housing demand. This
comparison helps in designing appropriate policies according to the
contemporary state of the housing market. Different factors contribute
differently towards housing demand on the face of two contrast economic
cycles.
Following the conventional housing studies, this study determines a
house price by employing the hedonic price model. Unlike most studies on
developing countries, this study quantified the relationship between the
housing demand and its covariates, by using an econometric framework,
augmented by Heckman's two-step selection procedure that controls
selectivity bias between the tenure choice and quantity of housing
services demanded. Margins for probability of house ownership are
calculated by using the Probit model. Permanent and Transitory income is
also estimated according to the permanent income hypothesis. Another
aspect, not commonly found in studies for developing countries,
including Pakistan, is determining the separate effect of permanent and
more importantly transitory income on housing demand. The log-linear
model is estimated using the Ordinary Least Square (OLS) technique.
The rest of the paper is organised as follows: section 2 presents
the state of housing in Pakistan with special focus on National Housing
Policy; section 3 layouts the conceptual framework of the study,
describes data and estimation methodology; section 4 explains the
results and last section concludes the discussion with policy
suggestions.
2. STATE OF HOUSING IN PAKISTAN
According to Population Census 1998, the stock of housing units was
19.2 million in 1998. Figure 1 represents the distribution of housing
units across the provinces. Figure 1 indicates that in 1998, 55 percent
housing units were in Punjab, 26 percent in Sindh, 11.5 percent in KPK
and 5.1 percent in Balochistan. The stock of housing units was 12.5
million in 1981.
The housing units, as a percentage of the total population remained
almost the same across provinces. The housing units as a percentage of
the total population have declined from 14.6 percent in 1981 to 12.5
percent in 1998 in KPK. The housing units as a percentage of the total
population have declined from 15.9 percent in 1981 to 14.3 percent in
1998 in Punjab. On the other hand, the housing units as a percentage of
the total population has increased from 14.6 percent in 1981 to 16.5
percent in 1998 in Sindh and from 13.6 percent in 1981 to 14.8 percent
in 1998 in Balochistan. The increase in housing units was primarily
observed in rural areas of Sindh and Balochistan during that period.
But, on the other hand, decline has been recorded in urban areas of
Sindh and Balochistan during this period (Table 1).
Table 2 presents the "nature of tenure" at national level
across the rural and urban areas. The nature of tenure was measured
using three categories, including "owned house", "rented
house" and "rent free house". The data uncovered that the
owned dwellings have increased from 78.4 percent in 1981 to 81.2 percent
in 1998. There was no significant change in the ratio of owned houses
from 1998 to 2012-13. The statistics have established that around 86
percent dwellings are owner occupied (Table 2). Similar patterns have
been observed across rural and urban areas of Pakistan. There was not a
huge change across rural and urban areas in the ratio of owned houses
from 1981 to 1998 (Table 2)
Various indicators are used to examine the level of congestions
within the housing unit. In this context we use persons per housing
unit, person per room, single room housing units, two rooms housing
units and three to four rooms housing units. Census of 1981 and 1998
established that in Pakistan persons per housing unit were 6.70 and 6.80
percent in 1981 and in 1998 respectively and the number of persons per
room was 3.50 and 3.13. On the other hand, it was noticed that 51.54 and
38.11 percent people were living in one room, whereas 44.83 and 30.54
percent, 3.63 and 24.43percent, 6.70 and 6.92 percent people were living
in two rooms, three to four rooms and five or more rooms respectively
(Table 3 and Table 4).
The gap between supply and demand for housing is persistently
rising in Pakistan. The previous section clearly indicates that per
annum housing demand is around 0.35 million. The unavailability of new
housing unit increases the congestion and homeless people in the
country. This calls for governmental intervention to provide decent
accommodation to every household. Pakistan had no housing policy at
national and even provincial level till 1992. First National Housing
Policy was developed in 1992, which was revised in 1994. This policy
proposed various innovative methods for increasing housing stock and
improving the quality of existing housing units. The government,
nevertheless, failed in implementing this policy. Later on, the
government of Pakistan had formulated a National Housing Policy (NHP) in
2001.
The NHP 2001 covers all major issues related to housing market,
such as land issues, housing finance, construction service sector, low
cost rural housing, building material, infrastructure development,
zoning regulations and institutional arrangements. The NHP 2001
highlights key challenges to housing sector and proposes some strategies
to resolve those issues and challenges and spells out the aims with key
objectives. Following are the key problems and issues that were
highlighted in the NHP 2001:
(i) The housing related issues are mainly generated by huge
population growth.
(ii) The per annum housing requirement is 0.57 million.
(iii) The unchecked growth of squatter settlements, Katchi Abadis,
encroachment of state and vacant land is held responsible for housing
shortages.
(iv) Scarcity of suitable land for housing, particularly in and
around the urban centres.
(v) Affordability issues, especially for low-income group.
(vi) The housing stock is rapidly aging.
(vii) Shortage of affordable housing finance to be major obstacle
in housing production.
(viii) Tremendous rise in price of housing material because of
inflationary pressure,
(ix) Lack of the use of technology.
To overcome these issues and meet the future housing requirements,
with low cost and high quality, the proposed NHP 2001 was intended to
achieve the following aims and objectives:
(i) To propose an enabling strategy for capacity building and
institutional arrangements.
(ii) Empowering all stakeholders, including public as well as
private sector for housing market development.
(iii) To propose a strategy for easing housing finance and home
improvement credits which are compatible with affordability, especially
for low-income group.
(iv) Strategy to improve the housing conditions through
development, capacity building and initiation of innovative ideas.
(v) Strategy to upgrade the existing cities with better planning
through the improvement of infrastructure.
(vi) Encourage research and development activities to design low
cost houses.
(vii) Provision of safeguard against malpractices and resource
mobilisation.
(viii) Provision of incentives through tax rationalisation.
(ix) A countrywide program of developing small and medium size
towns having growth potential.
3. CONCEPTUAL FRAMEWORK, DATA AND ESTIMATION METHODOLOGY
3.1. Conceptual Framework
To put the above discussion in a framework, we followed the work of
Goodman (1998), Zabal (2004) and Fontenla and Gonzalez (2009). These
studies have used the utility maximisation approach as a framework to
understand the housing demand dynamics. Let individual i's utility
function in market j depends on two goods: (i) non-housing composite
consumption, denoted by [C.sub.ij] and (ii) the amount of housing units,
denoted by [q.sub.ij]. We also assume that households have the same
utility function but differ in their socio-demographic characteristics.
These socio-demographic characteristics are denoted by [z.sub.i]. The
vector z includes variables such as age of the head of the household,
gender and education, social status and migration. The utility function
of the household can be written as follows:
[U.sub.ij] = U([C.sub.ij] [q.sub.ij] [z.sub.i]) (1)
Assuming a static setting, the objective of an individual is to
maximise the utility, given the budget constraints. An individual
chooses how to allocate his/her income to non-housing composite
consumption ([C.sub.ij]) and the housing services ([q.sub.ij]). The
budget constraint of an individual can be defined as follows:
[C.sub.ij] = [p.sub.i][q.sub.ij] = [m.sub.ij] (2)
Where m is the household's income, p is the price of housing
services and the price of non-housing consumption is normalised to one.
We allow housing prices to be different across markets.
The household's utility maximisation problem is defined as
follows:
[mathematical expression not reproducible] (3)
Solving the budget constraint for [C.sub.ij] and substituting into
the utility function gives the indirect utility function. The budget
constraint can be written as follows:
[C.sub.ij] = [m.sub.ij] - [p.sub.i][q.sub.ij] (4)
Now substituting Equation 4 in the utility function, we get the
following indirect utility function
[V.sub.ij] = [Max.sub.qij] U([m.sub.ij] - [p.sub.i][q.sub.ij],
[q.sub.ij], [z.sub.i]) (5)
Solving Equation (5) yields the (implicit) housing demand equation
[mathematical expression not reproducible] (6)
Providing a specific form for the utility function (1), will give
rise to an explicit housing demand equation. While many utility
functions result in non-linear demand equations, typically a log-linear
housing demand equation is specified
ln [q.sub.ij] = [[beta].sub.0] + [[beta].sub.1] ln p +
[[beta].sub.2] ln [z.sub.i] (7)
This equation can be assumed to be an approximation to the
underlying (nonlinear) housing demand equation. We analysed the housing
demand in Pakistan with the use of this model. Estimating the implicit
parameters of Equation (7) is the main purpose. We usually observe the
value of the housing unit rather than the quantity. Thus, [q.sub.ij] has
to be estimated in order to obtain Equation (7). An important feature of
the housing market is that the physical and surrounding characteristics
of the housing units are important, yet they vary widely across the
housing units.
Define [H.sub.n] as the vector that represents housing
characteristics for housing unit n. Similarly, [[beta].sub.j] is defined
as the parameter vector, which is allowed to vary across markets, for
each of the housing unit characteristics in [H.sub.n]. Thus the value y
of a housing unit n in market j, consumed by household i, is given by
the following expression:
[V.sup.i.sub.nj] = V([H.sub.n]; [[beta].sub.j]) ... ... ... ... ...
... ... (8)
If the characteristics [H.sub.n] and the value [v.sup.i.sub.n,j] of
each housing unit are known, then it is possible to estimate
[[beta].sub.j], using a hedonic price model. In addition, defining
[H.sup.*.sub.n] as the standard unit we can compute the price index
[p.sub.j] as follows:
[p.sub.j] = v([H.sup.*.sub.n]; [[beta].sub.j])/v([H.sup.*.sub.n];
[[beta].sub.1]) ... ... ... ... ... ... ... (9)
The value of the housing unit n in market j, consumed by the
household i can be expressed as [v.sup.i.sub.n,j] = [q.sub.iy] x
[p.sub.j]. The quantity, therefore, of housing is obtained as follows:
[q.sub.ij] = [v.sup.i.sub.nj]/[p.sub.j] ... ... ... ... ... ... ...
(10)
Once we know the [q.sub.ij], we can estimate the Equation 7.
3.2. Data
To estimate the demand for housing, data of various social and
economic indicators is taken from Pakistan Social and Living Standards
Measurement Survey (PSLM) and Household Integrated Economic Survey
(HIES), conducted by Pakistan Bureau of Statistics (PBS). In this study,
PSLM survey data for the year 2004-05 and 2010-11 is taken.
We are using a set of population based social indicators for 16341
households from PSLM/HIES national level data. The data on household
information covers education, health, employment and income as well as
ownership of assets, household details, immunisation, married women,
facilities and services. Additionally, it offers data on household
consumption expenditures (including consumption on durable items
owned/sold), transfer received and paid out and buildings and land
owned. Population of all the four provinces is considered as the
universal sample. Under the framework of PSLM/HIES each city/town was
sub divided into enumeration blocks. Urban areas were divided into 26698
blocks and rural areas comprised of 50588 blocks. Each urban block was
categorised on the basis of income groups. The selection of Primary
Sample Units (PSU) and Secondary Sample Units (SSUs) data from urban and
rural areas of each province has been discussed in Table 5.
Table 5 indicates that the entire sample of households has been
drawn from 1045 Primary Sample Units (PSUs) in 2004-05, out of which 486
are urban and 559 are rural and 1180 Primary Sample Units (PSUs) in
2010-11, out of which 564 are urban and 616 are rural. The total sample
is 14777 in 2004-05 and 16341 in 2010-11. This sample size has been
considered sufficient to produce estimates of key variables at national
and provincial levels [Pakistan (2012)].
3.3. Estimation Methodology
In this study, we seek to determine the factors that impact the
demand for housing and its services across income groups. Additionally,
welfare impact across income groups is also determined for 2004-05 and
2010-11, that will clarify whether the housing units owned by the
income-based groups are better off or worse off. For the purpose of
analysis, this model includes m = 1 ... M urban areas. In each urban
area there are i = 1 ... [I.sub.m] individual household heads and j = 1
... [J.sub.m] housing units. The analysis, therefore, considers each
city as a separate entity across income groups.
3.3.1. Housing Demand Model
The model explained in previous section indicates that various
socio-economic variables explain the housing demand. These factors
include different physical and community attributes, such as number of
rooms, dummy variable for owner occupied or rented unit as well as the
availability of housing services, including pipe-water, motors, hand
pumps or others. The community attributes include whether a housing unit
is located in a city, number of earners and educated members in the
household. Moreover, attributes related to head of the household are
age, gender, education, marital status, employment status and
occupation. In this study, we consider the household head as a special
case. Thus demand for housing can be represented as a function of
personal characteristic of the household head, background of the
household and price of housing. The functional form of housing demand
highlighted in Equation 7 at maximum utility level is given below:
q(z) = f([X.sub.1],Y,C,[P.sub.j](z)) ... ... ... ... ... ... (11)
For each housing characteristic "[z.sub.ij]" the q(z)
presents the quantity of individual housing unit, which is to be
estimated. [X.sub.1] refers to the household characteristics such as
family size, income group it belongs to and the number of earners.
Income of household is represented by Y which is the sum of permanent
income and transitory income. Characteristics of the head of the
household is represented by C, including education of the household
head, age of the household head, occupation, marital status and gender.
Finally, [p.sub.j](z) denotes the price of the individual household
which is not available for household data. A proxy, therefore, is used
to capture the price of the house estimated, using hedonic price model.
In order to estimate the demand for housing, we first need to calculate
the quantity of housing unit (q (2)), house price ([p.sub.j](z)),
permanent income ([T.sub.P]) and transitory income ([Y.sub.T]). Thus the
demand for housing is determined as
[Y.sub.i] = [x'.sub.i][beta] + [e.sub.i] ... ... ... ... ...
... ... (12)
In Equation (12), [Y.sub.i] represents the housing quantity and
[x'.sub.i] is a vector with dimensions 1 x M, representing all
exogenous variables included in the model, [beta] is a vector of
parameters with M x 1 dimensions. The following Ordinary Least Square
(OLS) regression equation specifications used for housing demand in
log-linear form is given as:
[mathematical expression not reproducible] (13)
Where, [beta] is the coefficient of exogenous variables and [delta]
is the coefficient used for dummies. The interpretation of the variables
is as: q(z) is the quantity of housing unit (defined in equation 10),
[Y.sub.P] is the permanent income of the household, [Y.sub.T] represents
the transitory income, [p.sub.j](z) is the price of housing unit,
[A.sub.R/Y] is the affordability of the individual household head, A is
the age of the household head, [E.sub.r] represents the number of
earners in a household, [E.sub.d] is the education of the household
head, [F.sub.s] is the family size (children and adults), male is the
dummy, representing the gender of the household head (=1 if male; = 0 if
female), M is the marital status (=1 if married; =0 otherwise),
[Y.sub.M] is the dummy for middle-income group (=1 if middle-income
group; = 0 otherwise), [Y.sub.H] is the dummy for high-income group (=1
if high-income group; = 0 otherwise), Ct represents the dummy for the 14
urban cities of Pakistan selected in this study and [e.sub.i] is the
error term. Moreover, all the variables used are in its log form for
reducing changes including extreme values in parametric estimation.
Additionally, it also reduces the heteroscedasticity in data.
3.3.2. Housing Quality
In order to estimate the housing demand, dependent variable i.e.
quality of housing units is first calculated following Dusansky and Koc
(2007).
Standardized Housing unit (q(z)) = (owner occupied housing
value)/house price per unit [p.sub.j] (14)
The market value of the house is used as a proxy to measure the
owner-occupied housing value, which refers to the price of the house
acceptable if he wishes to sell his property. The housing price per unit
represents the hedonic price [P.sub.j](z). Some studies also used rent
(rent equivalent) instead of owner-occupied housing value for the
calculation of housing units [Hernandez and Garcia (2006); Garabato and
Sarasola (2011)]. The demanded quantity of housing calculated in
Equation (14) is used to calculate the factors that affect housing
demand.
3.3.3. Permanent and Transitory Income
There are many views regarding the modelling of the unobservable
variables, such as permanent income and transitory income. Friedman
(1957) states that the consumption is the function of permanent income,
but his point of view was criticised as the consumption decision of the
household are forward-looking. It was looked at as a poor determinant to
measure the permanent income. Though, permanent income cannot be
measured directly, it is estimated using physical and human resources,
such as education, property and experience, which contribute in
generating income. Singh, et al. (1986) states that the determinants of
permanent income are the household characteristics, physical assets,
education, community and environmental attributes. It was, nonetheless,
argued that the physical assets are a weak determinant of permanent
income, as physical assets may underline a different level of permanent
income in different countries. Because of the environmental and economic
factors, the price of physical assets is distorted and it represents
different proportion of ownerships, thus the level of permanent income.
Many different approaches are discussed in literature to measure
permanent income, Townsend, et al. (1985). Some used qualitative
approach while others used rapid rural appraisal (RRA) approach
[Takasaki, et al. (2000)]. Shefer (1990), Ahmed (1994) and Ballesteros
(2001) used the expenditures on consumption as a proxy to measure
permanent income.
There are some studies, nevertheless, that used a set of different
individual characteristics such as education, age, skills, wages, bonus,
pension, on job training capital gain, inheritance and savings [Goodman
and Kawai (1984); Ahmed (1994); Wang (1995); Goodman (2002)]. Thus,
following Goodman and Kawai (1984) and Goodman (2002), we expressed
permanent income as a function of human and non-human wealth:
[Y.sub.P] = f(H,N) ... ... ... ... ... ... ... (15)
Where, H is the human wealth and N represents the non-human wealth.
The human wealth demands the expected future income such as bonus and
increments (annual increase in income on constant rate) and the current
income, which depends on the individual household characteristics such
as age, education, gender, marital status, occupation, employment
status, number of earners in a household and family size. Nonhuman
wealth accounts for the income received from other resources, such as
remittances and income from commercial or non- agricultural property.
Thus, the permanent income measure represents the potential
lifetime earnings and by regressing the real observed total income on
the independent variables, provides the permanent income as fitted value
of the regression and transitory income as residual. Observed total
income is indicated as the sum of permanent and transitory income is
highlighted in Equation (16) as:
Y = [Y.sub.p](H, N) + [Y.sub.T] ... ... ... ... ... ... (16)
Linear regression model is represented as:
[mathematical expression not reproducible] (17)
Where, [beta] is the coefficient of exogenous variables and [delta]
is the coefficient used for dummies. Y represents the observed total
income, A is the age of the household head, [E.sub.r] represents the
number of earners in a household, [E.sub.d] is the education of the
household head, [F.sub.s] is the family size (children and adults),
[R.sub.m] is the remittances, male is the dummy representing the gender
of the household head (=1 if male; = 0 if female), M is the marital
status (=1 if married; = 0 otherwise), [E.sub.p] is the dummy for
employment status (=1 if employed; = 0 otherwise), [O.sub.c] represents
the occupancy: whether the housing unit is owner occupied, rented,
subsidised rent or rent free, [Y.sub.M] is the dummy for middle-income
group (=1 if middle-income group; = 0 otherwise), [Y.sub.H] is the dummy
for high-income group (=1 if high-income group; = 0 otherwise), Ct
represents the dummy for the cities and [e.sub.i] is the error term.
Thus the predicted income is the required permanent income ([Y.sub.P])
and the residual is saved as transitory income ([Y.sub.T]).
Various income measures can be estimated using different sets of
explanatory variables, but the best fitted regression model for which
the standard error is minimum, is chosen for the analysis.
3.3.4. House Price
Since house prices are not available in the data set of PSLM, it is
estimated using the hedonic price model. The price of the house is
determined by the internal characteristics as well as the external
factors. There are other underlying issues that cause difficulties
towards calculating the price of unobserved variables. Firstly, price of
the property is not the same in each period; the house price varies
because of the supply and demand factors that determine the price. Thus,
the price is not same for two consecutive periods. Secondly, such as
many other products, properties of house traded in market are not
identical. The price changes, therefore, because of the characteristics
of property (number of rooms, appearance, source of water, availability
of gas, electricity, telephone, means of sewage), location attributes
(close to market area, office, school, hospital, neighbourhood and
others) and environmental attributes (urban, rural, industrial area, air
or water pollution) [Herath and Maier (2010)]. Thus, these attributes
cannot be ignored while calculating the house price.
The household survey data provides information about the expected
value of the house/property, if it is put up for sale and is reported as
the owner-occupied housing value. Owner-occupied housing value
represents the product of housing price per unit and standardised
housing unit. The value of the housing price is extracted from the
owner-occupied housing value using hedonic regression. Following Goodman
and Kawai (1984) and Goodman (2002), house price per unit can be
calculated using hedonic price model, which is a more sophisticated form
of mix adjustment. The hedonic regression, in terms of set of features
that contributes to the value of house is as follows:
[mathematical expression not reproducible] (18)
Where, P is the owner-occupied housing value, number of rooms is
used as a proxy to measure the house size and is represented by
[H.sub.S], [T.sub.H] is the house tax, [W.sub.P] is the dummy for piped
water (= 1 if piped water; = 0 otherwise), [W.sub.HP] is the dummy for
water from hand pump (= 1 if water from hand pump; = O otherwise),
[W.sub.M] represents the dummy for water availability from motor (= 1 if
water from motor; = 0 otherwise), [G.sub.AS] is the dummy for
availability of gas (= 1 if gas is available; = 0 otherwise), TS denotes
the time (in minutes) required to reach a grocery shop, TPT represents
the time (in minutes) required to reach a public transport, TPS
characterises the time (in minutes) required to reach a primary school,
TMS represents the time (in minutes) required to reach a middle school,
77/5 represents the time (in minutes) required to reach a high school,
THP symbolises the time (in minutes) required to reach a hospital,
[Y.sub.M] is the dummy for middle-income group (=1 if middle-income
group; = 0 otherwise), [Y.sub.H] is the dummy for high-income group (=1
if high-income group; = 0 otherwise), Ct represents the dummy for the 14
urban cities and e is the error term, [beta] is the coefficient of
exogenous variables and O is the coefficient used for dummies. Taking
logs of the variables are considered to ensure that the prices are
non-negative. This regression model used values of the above mentioned
features to predict the price of housing unit during a particular
period. The fitted values generated from the hedonic regression are the
required prices per house for the standardised housing units. Dusansky
and Koc (2007) are of the view that prices generated from the hedonic
method represent the prices of the same sized house across cities.
Hedonic price modelling is commonly used in real estate for sales
comparison. Thus, allow the comparison between prices of constant
quality housing across cities over a particular time period.
3.3.5. Imputed Rent
For the calculation of standardised unit of housing, we also need
to estimate affordability. In literature, affordability is defined as
the ratio of rent to total income [Tiwari and Parikh (1998)]. Housing
expense is commonly measured through rent. Malpezzi and Mayo (1985)
considered rent as the product of unit price and quantity consumed,
depending on the housing services. It varies for individual household,
depending on the shelter, type of construction, dwelling and
neighbourhood. The conventional hedonic regression model can also be
used to measure imputed rent. The hedonic equation for house rent is
specified as:
R = f([R.sub.T],[W.sub.T],[G.sub.AS],[T.sub.F],W) ... ... ... ...
... ... (19)
The house rent is measured against the set of characteristics of
the housing unit, which are specified as follows: [R.sub.T] is the type
of roof, it may be made of rcc/rbc, wood/bamboo, steel/cement or other;
[G.sub.AS] represent the dummy for availability of gas, [T.sub.F]
represents the toilet facility (outdoor, flush, pit/latrine or others),
W represents the water availability (piped, hand pump, motor or other)
and [W.sub.T] refers to type of walls i.e. brick, cement, stone, wood,
bamboo or mud. After the hedonic regress, fitted value of the imputed
rent is generated for the sample of housing units [Malpezzi (2003)].
Imputed rent is used only for owner-occupied housing unit for which only
market value of housing is available.
Following Greene (2003) and Wooldridge (2006), we applied
Heckman's twostep model [Heckman (1979)] of sample selection. In
order to select sample, two equations are used, first is the equation
that determines the outcome variable. Second equation only uses selected
samples and mechanisms determining the selection process. The dependent
variable, standardised housing unit, is only observed for those
household heads that are the owners of their houses and are not observed
for those rented units. Regressing an OLS model on the standardised
housing unit will cause sensitivity bias. Thus, the model will estimate
biased and incontinent value of [beta]. In order to deal with the
problem of non-random selection and to control the sensitivity bias
between household quantity of service demand and tenure, the Heckman
two-step selection model was applied.
For data generation, the Heckman model applies the moments of
incidentally reduced by variant normal distribution. The basic Heckman
model equation is specified as:
[mathematical expression not reproducible] (20)
The basic demand equation is
[mathematical expression not reproducible] (21)
[z.sup.*.sub.1] refer to those households who own their houses and
is 0 for rented households, [w.sub.i] is the lxk row vector for the
selected exogenous variables used in Heckman model and in the demand
equation, [gamma] is the parameter to be estimated with kxl dimensions.
As a special case, if the error terms of both equations are correlated
then the problem of selectivity arises and additional assumptions are
imposed:
[u.sub.i] ~ N(0,1) [[epsilon].sub.i] ~ N(0, [[sigma].sup.2])
corr([u.sub.i], [[epsilon].sub.i]) = [rho] ... ... ... ... ... ... ...
(22)
Here we assume normal distribution with mean zero and
correlation[rho]. Following Goodman (1988), Ahmad (1994) and Dusansky
and Koc (2007) the selectivity biasness was removed through Heckman
process.
4. RESULTS AND DISCUSSION
4.1. Housing Demand Analysis
The housing demand was estimated using 2-step Heckman model, whose
results are presented below in Tables 6 to 8. The housing demand during
2004-05 indicates that the coefficients of permanent income are
insignificant, whereas during 2010-11 the permanent income elasticity of
0.04, negatively yet significantly affected the housing demand. Whereas,
the transitory income elasticity 0.033 (2004-05) and 0.039(2010-11) was
found to be positive and significant. It was found to be relatively
smaller as compared to the income elasticity range i.e. 0.6 to 0.8 found
in literature for developing countries [Mayo (1981)]. The results are in
line with the findings of Malpezzi and Mayo (1987), Lodhi and Pasha
(1991) and Garabato and Sarasola (2011) and they indicate that
additional factors are needed to improve the housing demand. The results
in Table 5.1, nonetheless, present a static relationship between
transitory income and housing demand.
The difference in the results of income elasticity may have stemmed
from different income measures [Mayo (1981)]. Ahmad (1994) used the
permanent income, predicted from the income regression. Shefer (1990)
and Ballesteros (2001) used monthly household expenditures as an
indicator for permanent income. Arimah (1992) and Tiwari and Parikh
(1998) used total annual income as a proxy for permanent income.
Moreover, because of the difference in the data sample the results
uncovered the variations.
The fitted house price was statistically significant and it
highlighted a negative relationship between house price and demand for
both data sets. It indicated that with an increase in house price,
keeping other factors constant, the overall demand in housing market
will decrease. The reported price elasticity ranges 0.027and 0.042 for
2004-05 and 2010-11, respectively. Results suggest that the demand for
house price was inelastic. The range of price elasticity observed in
literature was from -0.1 to -0.9. The less elastic demand, nonetheless,
could be caused by limited supply of houses in markets. Consequently,
the household head was bound to pay the asking price of housing. Results
imply a downward sloping demand for housing service with no gain in
housing demand market.
Rent to income ratio, used as proxy for afford ability, derived
positive and significant results for the housing demand. With an
increase in affordability, which may have been triggered by the increase
in income or decrease in rent, the house demand had increased by 18.5
percent and 0.8 percent in 2004-05 and 2010-11, respectively. It implies
that household head was willing to buy a housing unit in order to avoid
the large housing expense.
All the demographic factors are found to be insignificant for
2004-05, whereas except gender all the demographic factors are
statistically significant for 2010-11. Household head's age implies
that the need for housing increases with the increase in age. With an
additional year in the age of household head, increases the housing
demand by 0.1 percent, nonetheless, after a certain age, as the children
move out, the demand for housing decreases (2010-11). It also implies
that with an increase in house demand because of age, there appears a
change in the taste of the individual [Goodman (1988); Fontenla, et al.
(2009)].
The coefficients for number of earners in a household were also
positive and significant. The value of earners coefficient indicates
that with an addition of one earning member, the housing demand
increased by 0:2 percent (2010-11). Similar results were also reported
by studies in Pakistan, such as Pasha and Ghaus (1990) and Nazli and
Malik (2003). It shows that 'a aingle earner (household head)
cannot save enough because of high consumption'expenses, therefore,
with an increase in number of earning members, the saving level
increases, which ultimately increases the demand for housing.
Highly significant results are reported about the effect of
education on housing demand. With an additional year in education, the
housing demand increased by 4.5 percent, which discovered that the
income of the household was in control. Such household could demand a
new housing unit with a change in its taste (2010-11). Additionally,
with an increase of one member in family, the demand for housing
increased, as the number of rooms was already assigned to the existing
members of the household. Thus, with an additional member, 0.3 percent
increase was recorded. For a household that belonged to middle-income
group, the demand increase was recorded as 2.4 percent and for a
high-income group, an increase of 8.4 percent in demand was reported
(2010-11), as shown in Table (5.3). Mixed results were reported in the
previous literature regarding the effect of demographic factors on
housing demand.
The LAMBDA coefficient indicates that the choice for housing was
made by considering the housing units consumption. The LAMBDA
coefficient was positive and significant for only 2010-11. In case of
Pakistan, this study established that a major increase in housing demand
was caused by the education factor.
4.2. Housing Demand Analysis at Disaggregated Level
This study estimated the housing demand, based on income groups for
owned housing and is reported in Table (9). The permanent income
elasticity was found positive and significant for low-income group and
high-income group, whereas the elasticity for middle-income group was
found inelastic (2004-05). The elasticity of 0.06 (low-income group) and
0.07 (high-income group) reported were higher than the results at the
national level but it still indicated that the housing demand was
inelastic. During 2010-2011, permanent income elasticity (0.04) was
negatively significant only for middle-income group. Transitory income
represents a positive and significant effect on the housing demand
across all income levels and for both data sets. With the increase in
income level, the housing demand became less elastic, as high-income
group represented an elasticity of 0.02.
Increase in house price negatively affect the housing demand for
all income groups. These results are highly significant for low-income
groups and the credibility decreases as we move from low-income group to
high-income group for both data sets. An increase of 36 percent was
observed in housing demand with the increase in affordability in
low-income group (2004-05). As we moved from low to high-income group,
affordability ratio decreased because of high-income and fixed rent cost
and, therefore, housing demand was less affected by the affordability
(2004-05, 2010-11). Moreover, during 2010-11 attribute of affordability
was insignificant for housing demand.
The coefficient of age of the household head was found to be
insignificant during 2004-05, across all income groups. The age factor
caused housing demand to change slightly by 0.1 percent for income
groups (2010-11). As seen in the results at the national level, age,
number or earners, household size (family members), education, gender
and marital status are statistically insignificant during 2004-05,
except the household size related to high-income group, which negatively
and significantly causes house demand to decrease by 0.5 percent.
Increase in the earning members of a household, positively affected
the house demand by 0.1 percent (low-income group) and 0.4 percent
(middle-income group), nevertheless, for a high-income group, the
housing demand decreases by 0.2 percent (2010-11). For middle-income
group, the increase in housing demand because of an additional year of
education was reported as 5.1 per cent higher as compared to other
groups. No significant impact of education, nonetheless, for high-income
group was found on the housing demand. For most of the time families
living in low-income groups and middle-income groups view housing as an
indicator for Social, Economic and Personal Security.
A logical increase in house demand was recorded with the increase
in family size for all income groups. Household head, being male,
negatively affect the housing demand in low-income group, whereas its
effect was statistically insignificant for other income levels. For a
married household, the house demand increased by 1.5 percent for a
low-income group, while results are insignificant for middle and
high-income groups. The results demonstrated a change in the attributes
of housing demand across income groups over the year. The housing
attributes for low-income groups highlighted that demand was more
sensitive to the change. Therefore, high proportion of income was spent
on the improvement and consolidation of housing units in low-income
group.
In an attempt to explain the differences of regional housing
demand, this study used the log-linear regression model for a
province-wise analysis of houses in urban Pakistan. The results are
presented in Table (10). The empirical results suggest that household
income was the significant factor, causing housing demand to vary among
province. The permanent income elasticity of 0.03 (2004-05) and 0.04
(2010-11) was highly significant and positive for Sindh. Similarly,
positive statistically significant and permanent income elasticity was
also reported for Punjab (2004-05) and Balochistan (2010-11) on housing
demand. Whereas, in case of KPK (2004-05) and Balochistan (2004-05), the
housing demand was negatively related to change in permanent income, but
the results are insignificant for Balochistan only (2004-05). The
results indicate that the demand for housing was inelastic for all
provinces.
Transitory income elasticity is positive and statistically
significant for four provinces, for both data sets, except for KPK
(2010-11). Moreover, the coefficient of transitory income presented
inelastic demand for housing across all regions. House price negatively
and significantly cause the demand for housing to decrease for all
regions except for KPK (2004-05) and Balochistan (2004-05). The result
implies that the housing demand was inelastic and relatively small for
all regions. The housing price was elastic for Balochistan (2010-11) and
was reported to be 0.8. It indicates that the sectorial and regional
difference should also be considered using aggregate parameters, while
estimating housing demands. These differences are not reflected in the
housing demand at national level.
Interesting results were reported regarding the affordability of a
household head across regions. Results indicated that households in
Punjab, Sindh and Balochistan depend on affordability (rent to income
ratio) for housing demand but, for KPK results were found insignificant.
With an increase in affordability ratio, the housing demand increased by
15 percent and 4 percent for Punjab; 19 percent and 0.4 percent for
Sindh and 12 percent and 14 percent for Balochistan for both data sets,
respectively. It indicates that individuals depend less on affordability
for housing demand. A decrease, therefore, in housing demand was
observed over the period for Punjab and Sindh. The dependence on income,
nevertheless, for house demand increased over the period for
Balochistan.
As discussed in the previous literature, the demographic factors
uncovered mixed results. With an increase in age, a significant change
in demand was recorded to the extent of Punjab (2010-11) and Balochistan
(2004-05). An increase in number of earners negatively and significantly
caused the decrease in demand by 0.9 percent for Sindh (2010-2011) and
insignificant for all the other regions. With an increase in education,
the demand for housing increased by 0.9 percent and 0.1 percent in
Punjab; 1.1 percent in Sindh (2010-11) and 5.8 percent for Balochistan
(2004-05). Household size negatively affected the demand by 0.3 percent
for Sindh (2010-11) and its impact was insignificant for other regions.
Similarly, gender and marital status were found irrelevant as a result
of regional demand analysis.
5. CONCLUSION AND POLICY IMPLICATIONS
The study has attempted to investigate the determinants of housing
demand in urban areas of Pakistan. The empirical analysis is carried out
using Pakistan Social and Living Standard Measurement (PSLM) survey
2004-05 and 2010-11. The hedonic price model is used for the estimation
of house prices. In order to control the selectivity bias between the
tenure choice and the quantity of housing services demanded,
Heckman's two-step selection procedure is used.
The empirical analysis shows that housing price and income
(temporary and permanent) play an important role in the determination of
the housing units' demand. An increase in houses' prices
causes decrease in demand for the housing units, while the housing
demand increases when the permanent income increases. It was found that
the transitory income has positive effect on the housing demand among
all three groups, but its impact was relatively stronger in case of
middle-income. Escalation of price was negatively related to demand, as
the housing demand decreases with the increase in the prices of housing
units. The affordability has the same effect on low and middle-income
groups, but for high-income group it was positive, yet reflects a lesser
value in terms of its coefficient. The demand for housing at national
level has positively impacted permanent income, as people with more
permanent income caused an increase in demand for houses. The scenario
changed in 2010-11 when people with more permanent income had a negative
impact on the demand for housing.
To manage rising housing demand, following policy implications
emerged from the empirical analysis:
(a) Analysis indicates that housing price has a negative impact on
housing demand. This finding suggests that government should focus on
developing effective and enforced price control mechanism. This will
help to control housing market hence reduce burden on city development.
(b) The economic development is one of the major determinants of
housing demand, as measured by income. Empirical analysis reveals that
rising income has a significant impact on new housing markets. To meet
the future demands with rising income, it is suggested that government
in collaboration with the private sector should develop new low cost
housing societies to cater future needs. Along with the economic
development, increasing urbanisation and population growth put pressure
on the housing sector. Government of Pakistan should devise national
housing policy on priority basis to address future housing demand in the
light of Vision 2025. This framework should address not only future
demands but also quality issues of housing sector, especially in mega
cities.
(c) Affordability is another important policy dimension of the
housing sector. The positive association between affordability and
housing demand implies that household prefers to purchase new housing
unit to manage housing expenses. A well functional rental market may
help to reduce housing pressure with innovative housing units. The
government should regulate rent market to manage rising demands.
In essence, government should design and implement new housing
policy to cater future need. The policy should consider future
development, house prices and affordability dimensions in its design.
The policy should also look into the institutional arrangement of this
sector to manage rental market and development of new societies,
especially in mega cities. The on-going census 2017 will provide a
golden opportunity to assess the demand and supply conditions of housing
market, in finalising the new housing policy. The Benazir Income Support
Programme (BISP) has also collected data on housing condition across
Pakistan, using National Socio-Economic Registry (NSER). This data can
also be used to understand the current housing situation in Pakistan. A
detailed study may be conducted using NSER and PSLM 2014-15 datasets to
review and suggest policy framework, keeping in view development,
housing price and affordability dimensions.
Appendix
Appendix Table 1
List of Explanatory Variables
Variables Description
Standardised housing unit Dependent variable
Permanent Income Monthly income, remittances or
wealth effect
Transitory Income Unexpected income, bonus
Remittances Total remittance was the sum of domestic
and foreign remittance
Market Value of House Price of owner occupied housing unit
Low-income Group Low income group
(as identified under PSU)
Middle-income Group Middle income group
(as identified under PSU)
High-income Group High income group
(as identified under PSU)
Affordability Affordability was defined as
Rent to Income Ratio
Household Head's Age Age of household head in year
Household Head's
Education Education of household head in year
Gender = 1 if Male; = 0 if female
Marital Status = 1 if married; = 0 otherwise
Occupation Post currently working on
Employment Status = 1 if employed; = 0 otherwise
Industry Industry in which households
head was working
Number of Earners Number of earners in a house hold
Family Size Number of members in a house hold
Household size Number of Rooms
Owner Occupied House = 1 if owner occupied, = 1 otherwise
Housing Expenditure
(rent) Rent in rupees
Imputed rent imputed rent was used for the missing
values of house rent
Type of roof It refers to the material used in roof
Type of walls It refers to the material used in
making walls
Water Facility water availability in house, piped,
motor water or other
City urban cities are chosen for analysis
Source: Author's own work.
Appendix Table 2
Means of Variables (2004-05)
Variables National Low-income
Standardised housing unit 1.00 1.00
Permanent Income 11.92 11.56
Transitory Income 0.00 0.00
Remittances 11026 5089
Market Value of House 1386344 661429
Affordability 0.85 0.83
Household Head's Age 46.24 44.71
Household Head's Education 9.84 8.65
Gender 0.93 0.95
Marital Status 0.89 0.91
Number of Earners 1.72 1.84
Household size 6.49 6.89
Owner Occupied House 0.71 0.69
Observation 2752 820
Variables Middle-income High-income
Standardised housing unit 1.00 1.00
Permanent Income 11.88 12.58
Transitory Income 0.00 0.00
Remittances 8808 26069
Market Value of House 922390 3735204
Affordability 0.85 0.87
Household Head's Age 45.93 49.43
Household Head's Education 9.60 11.63
Gender 0.94 0.90
Marital Status 0.89 0.87
Number of Earners 1.68 1.63
Household size 6.48 5.91
Owner Occupied House 0.70 0.74
Observation 1402 430
Source: Author's own Calculations.
Appendix Table 3
Means of Variables (2010-11)
Variables National Low-income
Standardised Housing Unit 1.00 1.00
Permanent Income 12.37 12.08
Transitory Income 0.00 0.00
Remittances 15712 8320
Market Value of House 2637309 1112030
Affordability 0.29 0.27
Household Head's Age 47.18 46.04
Household Head's Education 10.63 8.72
Gender 0.93 0.93
Marital Status 0.89 0.89
Number of Earners 1.70 1.78
Household Size 6.14 6.36
Owner Occupied House 0.68 0.70
Observation 3053 944
Variables Middle-income High-income
Standardised Housing Unit 1.00 1.00
Permanent Income 12.32 12.93
Transitory Income 0.00 0.00
Remittances 15381 27683
Market Value of House 2033660 6454766
Affordability 0.28 0.33
Household Head's Age 46.89 49.58
Household Head's Education 10.39 13.15
Gender 0.93 0.92
Marital Status 0.88 0.89
Number of Earners 1.73 1.51
Household Size 6.14 5.83
Owner Occupied House 0.68 0.66
Observation 1485 624
Source: Author's own Calculations.
Ayaz Ahmed, Nasir Iqbal, and Rehana Siddiqui
Ayaz Ahmed <
[email protected]> is Senior Research Economist,
Pakistan Institute of Development Economics, Islamabad. Nasir Iqbal
<
[email protected]> is Director Research, Benazir Income Support
Programme (BISP), Islamabad. Rehana Siddiqui <
[email protected]>
is Head, Department of Environmental Economics, Pakistan Institute of
Development Economics, Islamabad.
REFERENCES
Ahmed, I. (1994) Up to the Waist in Mud: Earth-based Architecture
of Rural Bangladesh. Dhaka: University Press Limited.
Arimah, B. C. (1992) An Empirical Analysis of the Demand for
Housing Attributes in a Third World Country. Land Economics 68:4,
366-379.
Ballesteros, Marife M. (2001) The Dynamics of Housing Demand in the
Philippines: Income and Lifecycle Effects. Philippine Institute for
Development Studies. (Discussion Papers DP 2001-15).
Dusansky, R. and C. Koc (2007) The Capital Gains Effect in the
Demand for Housing. Journal of Urban Economics 61:2, 287-298.
Fontenla, M. and F. Gonzalez (2009) Housing Demand in Mexico.
Journal of Housing Economics 18:1, 1-12.
Friedman, M. (1957) The Permanent Income Hypothesis: Theory of
Consumption Function. National Bureau of Economic Research 20-37.
Garabato, N. and M. R. Sarasola (2011) Are Uruguayan Housing
Policies Reaching the Poor? An Assessment of Housing Deficit, Housing
Informality and Usage of Housing Programs in Uruguay. Inter-American
Development Bank.
Goodman, A. C. (1988) An Econometric Model of Housing Price,
Permanent Income, Tenure Choice and Housing Demand. Journal of Urban
Economics 20, 327-353.
Goodman, A. C. (1998) Andrew Court and Invention of Hedonic Prices.
Journal of Urban Economics 44, 291-298.
Goodman, A. C. (2002) Estimating Equilibrium Housing Demand for
Stayers. Journal of Urban Economics, 1-24.
Goodman, A. C. and M. Kawai (1984) Replicative Evidence of the
Demand for Owner-occupied and Rental Housing. Southern Economic Journal
50:4, 1036-1057.
Greene, W. (2003) Simulated Maximum Likelihood Estimation of the
Normal-Gamma Stochastic Frontier Model. Journal of Productivity Analysis
19:2, 179-190.
Heckman, J. (1979) Sample Selection Bias as a Specification Error.
Econometrica 47:1, 153-161.
Herath, Shanaka, and Gunther Maier (2010) The Hedonic Price Method
in Real Estate and Housing Market Research. A Review of the Literature.
WU Vienna University of Economics and Business, Vienna. (SRE-Discussion
Papers, 2010/03).
Lodhi, A. and H. A. Pasha (1991) Housing Demand in Developing
Countries: A Case Study of Karachi in Pakistan. Urban Studie 28,623-634.
Malpezzi, S. and S. K. Mayo (1987a) The Demand for Housing in
Developing Countries. Economic Development and Cultural Change 35,
687-721.
Malpezzi, S. and S. K. Mayo (1987b) User Cost and Housing Tenure in
Developing Countries. Journal of Development Economics 25, 197-220.
Malpezzi, S. and S. Mayo (1985) Housing Demand in Developing
Countries. The World Bank, Washington, DC.
Mayo, S. K. (1981) Theory and Estimation in the Economics of
Housing Demand. Journal of Urban Economics 10:1, 95-116.
Nazli, H. and Sohail J. Malik (2003) Housing: Opportunity,
Security, and Empowerment for the Poor. The Pakistan Development Review
42:4, 893-908.
Oktay, E., A. Karaaslan, O. Alkan, and A. Kemal Celik (2014)
Determinants of Housing Demand in the Erzurum Province, Turkey.
International Journal of Housing Markets and Analysis 7:4, 586-602.
Pakistan, Government of (1981 and 1998) Population and Housing
Census of Pakistan. Islamabad: Pakistan Bureau of Statistics
Pakistan, Government of (2004-2005) Pakistan Social and Living
Standard Measurement Survey. Islamabad: Pakistan Bureau of Statistics.
Pakistan, Government of (2010-2011) Pakistan Social and Living
Standard Measurement Survey. Islamabad: Pakistan Bureau of Statistics.
State Bank of Pakistan (SBP) (2013) Annual Report of State Bank of
Pakistan (SBP). Karachi: State Bank of Pakistan.
Pasha, H. A. and A. Ghaus (1990) Magnitude of Housing Shortage in
Pakistan. The Pakistan Development Review 29:2, 137-153.
Pasha, H. A. and M. S. Butt (1996) Demand for Housing Attributes in
Developing Countries: A Study of Pakistan. Urban Studies 33:7,
1141-1154.
Saiz, A. (2010) The Geographic Determinants of Housing Supply. The
Quarterly Journal of Economics 125:3, 1253-1296.
Shefer, D. (1990) The Demand for Housing and Permanent Income in
Indonesia. Urban Studies 21, 259-272.
Singh, I., L. Squire, and J. Strauss (1986) Agricultural Household
Models: Extensions, Applications and Policy. Baltimore: Johns Hopkins
University Press.
Takasaki, Y., B. Barham, and O. Goomes (2000) Rapid-rural Appraisal
in Humid Tropical Forests: An Asset Possession-based Approach and
Validation Methods for Wealth Assessment Among Forest Peasant
Households. World Development 28:11, 1961-1977.
Tiwari, P. and J. Parik (1998) Affordability, Housing Demand and
Housing Policy in Urban India. Urban Studies 35:11, 2111-2129.
Townsend, P. (1985) A Sociological Approach to the Measurement of
Poverty: A Rejoinder to Professor Amartya Sen. Oxford Economic Papers
37, 659-668.
UN-Habitat (2016) World Cities Report 2016: Urbanisation and
Development Emerging Futures. United Nations Human Settlements Programme
(UN-Habitat), Nairobi, Kenya.
Wang, Y. P. (1995) Public Sector Housing in Urban China since 1949:
The Case of Xian. Housing Studies 10:1, 57-82.
Wooldridge, J. M. (2006) CLuster-Sample Methods in Applied
Econometrics. Cluster-Sample Methods in Applied Econometrics: An
Extended Analysis.
Zabal, J. E. (2004) The Demand for Housing Services. Journal of
Housing Economics 13:1, 16-35.
Table 1
Housing Units (TrendAnalysis)
All Areas Rural Urban
Region 1981 1998 1981 1998 1981 1998
Housing units (million)
Pakistan 12.51 19.21 9.01 13.18 3.50 6.03
KPK 1.61 2.21 1.38 1.84 0.23 0.37
Punjab 7.53 10.54 5.57 7.34 1.96 3.20
Sindh 2.78 5.02 1.56 2.85 1.22 2.17
Balochistan 0.59 0.97 0.50 0.78 0.09 0.20
Housing units as percent of total across the rural urban
Pakistan 100.0 100.0 72.0 68.6 28.0 31.4
KPK 100.0 100.0 85.7 83.3 14.3 16.7
Punjab 100.0 100.0 74.0 69.6 26.0 30.4
Sindh 100.0 100.0 56.1 56.8 43.9 43.2
Balochistan 100.0 100.0 84.7 79.9 15.3 20.1
Housing units as percent of total across the provinces
Pakistan 100.0 100.0 100.0 100.0 100.0 100.0
KPK 12.9 11.5 15.3 14.0 6.6 6.1
Punjab 60.2 54.8 61.8 55.7 56.0 53.1
Sindh 22.2 26.1 17.3 21.6 34.9 36.0
Balochistan 4.7 5.1 5.5 5.9 2.6 3.2
Persons per housing unit
Pakistan 6.73 6.89 6.70 6.78 6.83 7.14
KPK 6.87 8.03 6.81 8.01 7.39 8.08
Punjab 6.28 6.98 6.16 6.89 6.63 7.19
Sindh 6.84 6.06 5.26 5.47 8.85 6.84
Balochistan 7.32 6.76 7.40 6.40 7.78 7.85
Source: Population Census (1981, 1998).
Table 2
Nature of Tenure (Percentages) by Rural/Urban Areas
1981 1998
All Rural Urban All Rural Urban
Nature of Tenure Areas Areas
All Types 100 100 100 100 100 100
Owned 78.4 82.6 67.7 81.2 86.8 68.9
Rented 7.7 2.2 21.9 8.6 2.2 22.7
Rent Free 13.9 15.2 10.5 10.2 11.0 8.4
2004-05 2010-11
All Rural Urban All Rural Urban
Nature of Tenure Areas Areas
All Types 100 100 100 100 100 100
Owned 86.6 92.8 78.4 85.9 91.2 75.7
Rented 8.1 1.5 16.8 8.1 2.0 19.9
Rent Free 5.3 5.7 4.8 6.0 6.8 4.4
Source: Population Census (1981, 1998);
PSLM (2004-05 and 2010-11).
Table 3
Congestion of Housing Units
1981
Indicators Pakistan KPK Punjab
Persons per Housing Unit 6.70 7.00 6.50
Persons per Room 3.50 3.60 3.30
Single Room Housing Units (%) 51.54 50.00 48.00
Two Rooms Housing Units (%) 44.83 4.00 48.00
3-4 Rooms Housing Units (%) 3.63 46.00 4.00
5 and more Rooms Housing Units (%) 6.70 7.00 6.50
1998
Persons per Housing Unit 6.80 8.00 6.90
Persons per Room 3.13 3.34 3.04
Single Room Housing Units (%) 38.11 27.71 31.97
Two Rooms Housing Units (%) 30.54 34.50 33.54
3-4 Rooms Housing Units (%) 24.43 29.11 27.12.
5 and more Rooms Housing Units (%) 6.92 8.67 7.36
1981
Indicators Sindh Balochistan
Persons per Housing Unit 7.10 7.60
Persons per Room 4.00 4.20
Single Room Housing Units (%) 61.00 60.00
Two Rooms Housing Units (%) 36.00 36.00
3-4 Rooms Housing Units (%) 3.00 4.00
5 and more Rooms Housing Units (%) 7.10 7.60
1998
Persons per Housing Unit 6.00 6.70
Persons per Room 3.37 3.07
Single Room Housing Units (%) 56.93 42.77
Two Rooms Housing Units (%) 23.87 25.18
3-4 Rooms Housing Units (%) 17.00 22.69
5 and more Rooms Housing Units (%) 3.56 9.36
Source: Population Census (1981 and 1998).
Table 4
Congestion of Housing Units
2004-05
Indicators Pakistan KPK Punjab
Persons per Housing Unit 6.75 7.71 6.55
Single Room Housing Units (%) 24.20 18.35 24.75
2-4 Rooms Housing Units (%) 68.71 69.90 68.69
5 and more Rooms Housing Units (%) 7.09 11.75 6.56
2010-11
Persons per Housing Unit 6.38 7.17 6.16
Single Room Housing Units (%) 24.83 19.03 26.09
2-4 Rooms Housing units (%) 69.33 72.62 67.49
5 and more Rooms Housing units (%) 5.84 8.32 6.43
2004-05
Indicators Sindh Balochistan
Persons per Housing Unit 6.71 6.88
Single Room Housing Units (%) 30.76 14.79
2-4 Rooms Housing Units (%) 65.00 75.78
5 and more Rooms Housing Units (%) 4.24 9.43
2010-11
Persons per Housing Unit 6.39 7.08
Single Room Housing Units (%) 25.67 20.89
2-4 Rooms Housing units (%) 70.94 75.02
5 and more Rooms Housing units (%) 3.39 4.09
Source: PSLM (2004-05 and 2010-11).
Table 5
Profile of the Sample of PSLM Survey (2004-05 and 2010-11)
Sample PSUs Sample SSUs
Province/Area Urban Rural Total Urban Rural Total
2010-11
Punjab 256 256 512 2935 4019 6954
Sindh 152 144 296 1802 2296 4098
Khyber Pakhtunkhwa 88 120 208 1041 1913 2954
Balochistan 68 96 164 811 1524 2335
Total 564 616 1180 6589 9752 16341
2004-05
Punjab 210 226 436 2511 3607 6118
Sindh 125 125 250 1497 1980 3477
Khyber Pakhtunkhwa 91 118 209 1088 1878 2966
Balochistan 60 90 150 713 1434 2147
Total 486 559 1045 5809 8899 14708
Source: PSLM/HIES (2004-05 and 2010-11).
Table 6
Tenure Choice Regression 2004-05 (Income Group Wise)
Low-income
Variables Coefficient Margin effect
Permanent Income 13.030 3.246
(1.01) *** (0.30) ***
Transitory Income -0.150 -0.037
(0.18) (0.05)
House Price 0.823 0.205
(0.33) ** (0.08) **
Affordability 5.162 1.286
(1.53) *** (0.39) ***
Age -0.089 -0.022
(0.01) *** (0.00) ***
Number of Earners -1.180 -0.294
(0.11) *** (0.03) ***
Head's Education -5.826 -0.893
(0.46) *** (0.03) ***
Household Size -0.436 -0.109
(0.04) *** (0.01) ***
Gender 1.340 0.468
(0.39) *** (0.15) ***
Marital Status -1.113 -0.168
(0.33) *** (0.03) ***
Intercept -152.635
(11.49) ***
City Dummies Yes Yes
No. of Observations 784 784
Middle-income
Variables Coefficient Margin effect
Permanent Income 8.211 2.405
(0.53) *** (0.17) ***
Transitory Income -0.244 -0.071
(0.10) ** (0.03) **
House Price 0.962 0.282
(0.19) *** (0.05) ***
Affordability 4.087 1.197
(1.10) *** (0.32) ***
Age -0.048 -0.014
(0.01) *** (0.00) ***
Number of Earners -0.708 -0.207
(0.06) *** (0.02) ***
Head's Education -3.709 -0.541
(0.26) *** (0.03) ***
Household Size -0.271 -0.079
(0.03) *** (0.01) ***
Gender 0.730 0.257
(0.22) *** (0.09) ***
Marital Status -0.064 -0.018
(0.16) (0.05)
Intercept -106.222
(6.16) ***
City Dummies Yes Yes
No. of Observations 1,402 1,402
High-income
Variables Coefficient Margin effect
Permanent Income 6.450 1.659
(0.77) *** (0.22) ***
Transitory Income -0.241 -0.062
(0.14) * (0.04) *
House Price 1.571 0.404
(0.27) *** (0.06) ***
Affordability 2.162 0.556
(1.55) (0.40)
Age -0.034 -0.009
(0.01) *** (0.00) ***
Number of Earners -0.569 -0.146
(0.11) *** (0.03) ***
Head's Education -2.993 -0.243
(0.46) *** (0.03) ***
Household Size -0.293 -0.075
(0.05) *** (0.01) ***
Gender 0.832 0.273
(0.37) ** (0.14) *
Marital Status -0.702 -0.137
(0.36) ** (0.05) ***
Intercept -99.540
(9.58) ***
City Dummies
No. of Observations 485 485
Note: *** p<0.01, ** p<0.05, * p<0.1, Figures
in parenthesis are robust standard errors.
Table 7
Tenure Choice Regression 2010-2011 (Income Group Wise)
Low-income
Variables Coefficient Margin Effect
Permanent Income 1.264 0.405
(0.47) *** (0.15) ***
Transitory Income -0.396 -0.127
(0.13) *** (0.04) ***
House Price 1.781 0.570
(0.25) *** (0.08) ***
Affordability 0.638 0.204
(0.32) ** (0.10) **
Age -0.001 -0.000
(0.01) (0.00)
Number of Earners -0.196 -0.063
(0.07) *** (0.02) ***
Head's Education -0.717 -0.213
(0.24) *** (0.07) ***
Household Size -0.079 -0.025
(0.03) ** (0.01) **
Gender 0.784 0.290
(0.26) *** (0.10) ***
Marital Status -0.632 -0.166
(0.22) *** (0.05) ***
Intercept -38.854
(5.81) ***
City Dummies Yes Yes
No. of Observations 944 944
Middle-income
Variables Coefficient Margin Effect
Permanent Income 1.391 0.465
(0.41) *** (0.14) ***
Transitory Income -0.176 -0.059
(0.09) ** (0.03) **
House Price 1.175 0.393
(0.17) *** (0.06) ***
Affordability 1.435 0.480
(0.28) *** (0.09) ***
Age 0.003 0.001
(0.00) (0.00)
Number of Earners -0.253 -0.085
(0.05) *** (0.02) ***
Head's Education -0.888 -0.251
(0.21) *** (0.05) ***
Household Size -0.016 -0.005
(0.03) (0.01)
Gender 0.474 0.174
(0.19) ** (0.07) **
Marital Status -0.337 -0.103
(0.15) ** (0.04) **
Intercept -33.092
(4.85) ***
City Dummies Yes Yes
No. of Observations 1,484 1,484
High-income
Variables Coefficient Margin Effect
Permanent Income 3.546 1.142
(0.71) *** (0.24) ***
Transitory Income -0.206 -0.066
(0.12) * (0.04) *
House Price 2.113 0.681
(0.24) *** (0.07) ***
Affordability 0.730 0.235
(0.21) *** (0.07) ***
Age -0.008 -0.002
(0.01) (0.00)
Number of Earners -0.475 -0.153
(0.10) *** (0.03) ***
Head's Education -2.362 -0.340
(0.41) *** (0.03) ***
Household Size -0.137 -0.044
(0.05) *** (0.02) ***
Gender 0.361 0.126
(0.30) (0.11)
Marital Status -0.415 -0.118
(0.27) (0.07) *
Intercept -74.003
(9.19) ***
City Dummies Yes Yes
No. of Observations 624 624
Note: *** p<0.01, ** p<0.05, * p<0.1, Figures
in parenthesis are robust standard errors.
Table 8
Housing Demand at National Level (Dependent Variable Housing Units)
Variable 2004-05 2010-11
Permanent Income 0.0190 -0.0430
(0.0200) (0.01) ***
Transitory Income 0.0330 0.0390
(0.00) *** (0.00) ***
House Price -0.0270 -0.0420
(0.00) *** (0.01) ***
Affordability 0.1850 0.0080
(0.02) *** (0.00) ***
Age 0.0000 0.0010
0.0000 (0.00) ***
Number of Earners -0.0010 0.0020
0.0000 (0.00) ***
Head's Education 0.0080 0.0450
(0.0100) (0.01) ***
Household Size -0.0010 0.0030
0.0000 (0.00) ***
Gender -0.0070 -0.0070
(0.0100) (0.01)
Marital Status 0.0000 0.0060
0.0000 (0.00) ***
Middle-income Group 0.0030 0.0240
(0.0100) (0.01) ***
High-income Group 0.0110 0.0840
(0.0200) (0.02) ***
Lambda -0.0020 -0.0180
0.0000 (0.01) *
Intercept 0.9820 2.0370
(0.23) *** (0.23) ***
City Dummies Yes Yes
Number of Observations 2752 3,040
R-squared 0.107 0.119
Note: *** p<0.01, ** p<0.05, * p<0.1, Figures
in parenthesis are robust standard errors.
Table 9
Housing Demand by Income Group (Dependent Variable Housing Unit)
2004-05
Variables Low-income Middle-income High-income
Permanent Income 0.0600 0.0240 0.0780
(0.04) * (0.0300) (0.04) *
Transitory Income 0.0360 0.0270 0.0280
(0.01) *** (0.00) *** (0.01) ***
House Price -0.0450 -0.0270 -0.0100
(0.01) *** (0.01) *** (0.0100)
Affordability 0.3620 0.1510 0.2410
(0.04) *** (0.04) *** (0.06) ***
Age 0.0000 0.0000 0.0000
(0.000) (0.000) (0.000)
Number of Earners -0.0060 -0.0030 0.0010
(0.000) (0.000) (0.000)
Head's Education (0.0090) 0.0050 (0.0200)
(0.0200) (0.0100) (0.0200)
Household Size -0.0010 0.0000 -0.0050
0.0000 0.0000 (0.00) **
Gender (0.0110) 0.0070 (0.0110)
(0.0100) (0.0100) (0.0100)
Marital Status (0.0040) (0.0020) (0.0080)
(0.0100) (0.0100) (0.0100)
Lambda 0.0050 (0.0040) 0.0310
(0.0100) (0.0100) (0.01) **
Intercept 0.6310 0.9640 0.0240
(0.4100) (0.32) *** (0.5800)
City yes Yes yes
No. of Observations 784 1402 485
R-squared 0.2390 0.0980 0.2640
2010-2011
Variables Low-income Middle-income High-income
Permanent Income (0.0240) -0.049 -0.009
(0.0200) (0.02) *** (0.04)
Transitory Income 0.0320 0.042 0.02
(0.00) *** (0.00) *** (0.00) ***
House Price -0.0410 -0.054 -0.019
(0.01) *** (0.01) *** (0.01) *
Affordability 0.0040 0.004 0.01
(0.00) *** (0.00) *** (0.01)
Age 0.0010 0.000 0.001
(0.00) *** (0.00) *** (0.00) ***
Number of Earners 0.0010 0.004 -0.002
(0.00) *** (0.00) *** (0.00) ***
Head's Education 0.0310 0.051 0.028
(0.01) *** (0.01) *** (0.02)
Household Size 0.0010 0.002 0.002
(0.00) *** (0.00) ** (0.00) ***
Gender -0.0210 -0.002 -0.012
(0.01) ** (0.01) (0.01)
Marital Status 0.0150 0.002 0.011
(0.01) * (0.01) (0.01)
Lambda (0.0250) -0.045 0.028
(0.01) * (0.01) *** (0.01) **
Intercept 1.8210 2.333 1.307
(0.32) *** (0.25) *** (0.57) **
City yes Yes Yes
No. of Observations 944.0000 1,484 612
R-squared 0.2030 0.16 0.297
Note: *** p<0.01, ** p<0.05, * p<0.1, Figures
Table 10
Housing Demand: A Provincial Comparison (Dependent Variable Housing
Unit)
Punjab Sindh
Variables 2004-05 2010-11 2004-05 2010-11
Permanent Income 0.014 0.000 0.038 0.044
(0.01) * (0.01) (0.01) *** (0.01) ***
Transitory Income 0.025 0.031 0.045 0.047
(0.00) *** (0.00) *** (0.00) *** (0.00) ***
House Price -0.021 -0.017 -0.017 -0.011
(0.00) *** (0.00) *** (0.01) *** (0.00) **
Affordability 0.152 0.044 0.199 0.004
(0.03) *** (0.01) *** (0.05) *** (0.00) *
Age 0.000 0.000 -0.000 0.000
(0.00) (0.00) * (0.00) (0.00)
No. of Earners 0.000 -0.000 -0.002 -0.009
(0.00) (0.00) (0.00) (0.00) ***
Head's Education 0.009 0.014 0.004 0.011
(0.00) * (0.00) *** (0.01) (0.01) *
Household Size -0.000 0.001 -0.001 -0.003
(0.00) (0.00) (0.00) (0.00) ***
Gender -0.009 0.001 -0.003 -0.005
(0.01) (0.01) (0.01) (0.01)
Marital Status 0003 -0.001 -0.004 0.001
(0.01) (0.00) (0.01) (0.01)
Intercept 0988 1.200 0.614 0.636
(0.05) *** (0.05) *** (0.07) *** (0.07) ***
No. of Observations 1,385 1,542 971 1,089
R-squared 0.066 0.106 0.144 0.247
KPK Balochistan
Variables 2004-05 2010-11 2004-05 2010-11
Permanent Income -0.047 0.023 -0.035 0.074
(0.02)** (0.02) (0.03) (0.03) **
Transitory Income 0.017 0.008 0.033 0.063
(0.01)** (0.01) (0.01) *** (0.01) ***
House Price 0.002 -0.030 -0.014 -0.087
(0.01) (0.01) ** (0.02) (0.01) ***
Affordability 0.071 -0.005 0.124 0.144
(0.10) (0.01) (0.07) * (0.04) ***
Age 0.000 0.000 0.001 0.000
(0.00) (0.00) (0.00) ** (0.00)
No. of Earners -0.001 -0.005 0.003 -0.008
(0.00) (0.00) (0.01) (0.01)
Head's Education 0.023 -0.008 0.058 0.017
(0.01) (0.01) (0.01) *** (0.02)
Household Size 0.001 0.001 -0.001 -0.001
(0.00) (0.00) (0.00) (0.00)
Gender -0.011 0.022 0.034 0.061
(0.02) (0.02) (0.03) (0.05)
Marital Status 0.016 -0.019 -0.016 -0.016
(0.02) (0.01) (0.02) (0.02)
Intercept 1.451 1.127 1.394 1.218
(0.15)*** (0.18) *** (0.21) *** (0.27) ***
No. of Observations 234 231 162 178
R-squared 0.126 0.078 0.263 0.391
Note: *** p<0.01, ** p<0.05, * p<0.1 level of
significance and standard errors are
highlighted in parentheses.
Fig. 1. Distribution of Housing Units across Provinces
KPK; 11.5
Punjab; 54.8
Sindh; 26.1
Balochistan; 5.1
Source: Authors' own calculation base on the "Population Census
1998".
Note: Table made from bar graph.
COPYRIGHT 2018 Reproduced with permission of the Publications Division, Pakistan Institute of Development Economies, Islamabad, Pakistan.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.
