COMPARISON OF URBAN HOUSING SATISFACTION IN
MODERN AND TRADITIONAL NEIGHBORHOODS IN
EDIRNE, TURKEY
(Accepted 14 March 2006)
ABSTRACT. Numerous cross-cultural studies have focused on certain aspects of urban
housing conditions and their social consequences. However, most data on housing satisfaction
is restricted to Western countries. Relatively little comparison has been made between these
findings and those in developing areas where rapid urbanization is occurring and where con-
comitant problems in urban housing are emerging. Moreover, since primary cities of developing
countries in the initial stage of economic development have received extensive attention in
urban housing research, it would be interesting to examine a non-primary city where relatively
good standards of living have been achieved. Thus, this study investigates people’s housing
satisfaction in modern and historical neighborhoods. This paper addresses some conceptual and
measurement issues related to the study of housing satisfaction. We build a conceptual model,
which links the multiple dimensions of housing satisfaction, measured by a modified version of
Bardo and Dokmeci’s (1992, Genetic, Social and General Psychology Monographs 118(3))
housing satisfaction scale, in a causal sense. An empirical examination of the model in tradi-
tional and modern neighborhoods reveals that social and environmental living conditions
positively influence overall housing satisfaction. The results also indicate that the drivers of the
social and environmental living conditions constructs differ between traditional and modern
neighborhoods.
KEY WORDS: Edirne, housing, satisfaction, measurement invariance, multi-group analysis
1.
INTRODUCTION
Edirne, located near the Greek-Turkish border of Turkey, is a historical
capital of the Ottoman Empire and a university center. Throughout its
history it has been a significant business and education center of Thrace.
Recently, while some modern districts have gained comparative advantage,
the traditional ones have started to decrease in population due to deterio-
ration of urban environments. Thus, districts of Edirne and their neigh-
borhoods are undergoing continuous social, economic and structural
transformation as a result of local and global pressures. The purpose of this128
NEVNIHAL ERDOGAN ET AL.
paper is to compare residential satisfaction between modern and old
neighborhoods in Edirne, to determine the main drivers of residential sat-
isfaction, and search for discrepancies between the two processes.
The present study contributes to the academic body of knowledge in
several ways. We first explore the nature of housing satisfaction then by
unifying the existing theories in the literature we develop a conceptual model
of housing satisfaction. Finally we test the model empirically. From a
practical point of view, we argue that the proposed multidimensional scale
and the model in this study can be used by the local authorities as a diag-
nostic tool to identify areas where specific improvements are needed and to
pinpoint aspects of housing satisfaction that require attention. Local
authorities such as housing policy makers and urban planners may use this
framework to develop relevant and effective strategies and to improve the
conditions that cause dissatisfaction in the new and peripheral areas.
This article has sought to improve understanding of the determinants of
housing satisfaction among the residents of modern (Murat I) and tradi-
tional (Kaleici) districts. Identifying the determinants and degree of housing
satisfaction provides important insights. Policy makers, local authorities
and the Municipality can use this information to coordinate housing goals
and to develop relevant and effective strategies.
The paper is organized as follows. Next section briefly reviews the liter-
ature on housing satisfaction and introduces a conceptual model of housing
satisfaction. Section Empirical analysis, outlines the measurement instru-
ment, sampling procedure, measurement and structural models used in the
analysis. Section Findings summarizes the findings and the last section
concludes.
2.
LITERATURE REVIEW OF HOUSING SATISFACTION
In recent years, one major aim of city planners has been to prevent the
deterioration of urban environments and thus stimulate quality develop-
ment of cities. Central to this development, research aimed at exploring the
relationship between people and their everyday urban environments has
increased. Understanding the individual–environment relationship and the
congruence or dissonance between the city dweller and his urban sur-
roundings is the quintessential planning problem towards understanding
(Michaelson, 1977; Rapoport, 1985). Over recent decades, considerable ef-
fort has been directed toward assessing the quality of different residential
environments (Pacione 1990; Bonaiuto and Aiello, 1999). Collectively this
line of research has contributed valuable insights into such questions as theCOMPARISON OF URBAN HOUSING SATISFACTION
129
extent and distribution of substandard housing and of deprivation in the
modern city (Pacione, 1986).
Residential satisfaction is a complex term as its precise meaning depends on
the place, time, and purpose of the assessment and on the value system of the
assessor, involving an extensive range of people: architects, planners, soci-
ologists, psychologists and urban geographers (Bardo and Dokmeci, 1992).
Previous research took various personal, physical, demographic and
social characteristics into consideration in studying residential satisfaction,
such as length of residence (Kasarda and Janowitz, 1974; Goudy, 1977,
1982; Hunter, 1978, 1979; Newman and Duncan, 1979; St. John et al., 1986;
Satsangi and Kearns, 1992), socio-economic status (Marans and Rodgers,
1975; St. John and Clark, 1984), and age (Marans and Rodgers, 1975;
Goudy, 1982; Barrasi et al., 1984). Physical structure and the physical
environment also appear to play a role in community satisfaction (Wirth,
1938; Guest and Lee, 1983; Bardo and Dokmeci, 1990, 1992).
Personal factors may also affect residential satisfaction, including previ-
ous housing experience (Fried and Gleicher, 1961), the degree of integration
into society (Tauber and Levin, 1971), the reference group (Merton, 1968),
and the socio-psychological attitude toward society in general (Gans, 1967),
and people’s social customs and traditions (Duncan, 1971).
Green areas and access to services and facilities and their quality are
found to be related to residential satisfaction (Marans and Rogers, 1975).
According to Duncan (1971), some families have no need for a garden while
others enjoy tending a fair-sized green area. Some wish to live close to a
town center for convenience; others do not mind a journey to work if they
can live in more open surroundings (Pacione, 1990).
In addition to the characteristics of the house, neighborhood, and resi-
dent, the habitability of a residential setting can be affected by city man-
agement (Stamps, 1994); for instance, the standard of garbage collection
and other local services (Onibokum, 1974).
The centrality of the residential environment for individual quality of life
has been established (Altman and Werner, 1985; Altman and Wandersman,
1987; Francescate et al., 1987). The quality of the residential environment
can be investigated in two ways. The first method uses such objective
measures as the number and range of facilities available (Wesserman, 1982)
and housing unit characteristics. The second method involves making sub-
jective assessments of resident satisfaction with their housing situations
(Weideman and Anderson, 1985).
Residential-location preferences with respect to different age groups,
household sizes, and income groups reveal that younger individuals’130
NEVNIHAL ERDOGAN ET AL.
preferences are concentrated in the periphery, while a large percentage of
middle and older age groups prefer to move to the intermediate area be-
tween the core and the periphery, now the most easily accessible zone in the
city (Dokmeci and Berköz, 2000).
Unifying the constructs reviewed in the preceding section, we propose the
conceptual model shown in Figure 1. We argue that the overall housing
satisfaction (OHS) is directly influenced by perceived living conditions (LC),
while perceived LC are related to satisfaction with the physical surrounding
(PS), satisfaction with the social relations (SR), satisfaction with the per-
formance of the local authorities (LA), and perceived quality of the facilities
(FQP). Moreover we allow satisfaction with the performance of the local
authorities to have an indirect effect on the perceived living conditions
through perceived quality of the facilities. Next we test the proposed model
using survey data from two neighborhoods characterized by traditional and
modern backgrounds.
3.
3.1.
EMPIRICAL ANALYSIS
Sample
Based on random starts, systematic samples are drawn from two districts of
Edirne, namely Kaleici (n = 114) and Murat I (n = 120). Edirne’s tradi-
tional district Kaleici has a historical past (accumulation of different culture)
from Roman, Byzantine, and Ottoman period to the date. Kaleici has been
redeveloped after being ruined as a result of a fire in the beginning of 20th
century. Although very affluent people lived in the Kaleici area in the past,
less well of people have predominated in the last 20 years. In contrast,
Fig. 1. Conceptual model. Note: OHS – overall housing satisfaction, LC – perceived living
conditions, PS – satisfaction with the physical surrounding, SR – satisfaction with the social
relations, LA – satisfaction with the performance of the local authorities, FQP – perceived
quality of the facilities.COMPARISON OF URBAN HOUSING SATISFACTION
131
Murat I is a newly developing area with apartment blocks, and the people
living in this district are higher in terms of education and social status
compared to traditional districts such as Kaleici. Murat I is a planned,
modern district on the periphery of the city, while Kaleici is located centrally
and considered as a traditional district. Residents of Murat I tend to be
more middle-class, while Kaleici is populated by traditional working class
residents. Recently, Kaleici has begun to experience urban gentrification and
the displacement of the traditional population (See photos of Kaleici and
Murat I).
3.2.
Data Collection Instrument
The urban housing satisfaction scale used in this article is initially developed
by Bardo and Dokmeci (1992). We use a modified version of the scale,
which appears to be multi-dimensional. A thorough investigation of the 47
items listed in the Appendix, reveals a theoretical multi-dimensional struc-
ture, compromised of dimensions (35 items) such as:
•
•
•
•
•
overall housing satisfaction (OHS – 8 items),
perceived living conditions (LC – 6 items),
satisfaction with the physical surrounding (PS – 6 items),
satisfaction with the social relations (SR – 10 items),
satisfaction with the performance of the local authorities (LA – 5
items).
In addition, we include a construct measuring the perceived FQP available
in the neighborhood. FQP is a composite measure, which is obtained by
weighing the quality perceptions of the residents for the available facilities in
the neighborhood with their visiting frequencies. Later, a FQP value, which
is comparable between subjects, is calculated by standardizing it to per visit,
per facility basis.
Prior to data collection, group discussions were conducted to ensure that
the items are also valid for the selected province of Turkey. After making
modifications, if necessary, in the wording and/or the content of the items, a
survey including the modified urban housing satisfaction scale, the battery
for facility quality perception, and a set of questions on demographics is
administered in two distinct districts of Edirne, Turkey. Items measuring
urban housing satisfaction are scored on a five-point scale, ranging from
strongly disagree (=1) to strongly agree (=5).132
3.3.
NEVNIHAL ERDOGAN ET AL.
Measurement Model
In this section, we describe the scale for the measurement of urban housing
satisfaction and other related constructs, as well as the building blocks of the
modified models that allow us to compare housing satisfaction in modern
and traditional neighborhoods.
Initially, we test whether the items (1) represent their hypothesized
components, as reflected in high loadings on the corresponding component,
(2) do not confound the multiple components that are defined earlier, as
reflected in low cross-loadings, and (3) cover the multiple components in as
many different shades as possible. At this stage, we perform all analyses on
the two samples separately. First, item-to-total correlations are computed
for the given satisfaction items, and items that do not correlate significantly
better with the hypothesized than the non-hypothesized component are
eliminated (DeVellis, 1991). In general items belonging to LC, PS, and SR
dimensions exhibit lower correlations with the hypothesized constructs than
the items in the remaining dimensions. Analyses based on item-to-total
correlations leave us with eight OHS items, six LC items, six PS items, ten
SR items, and five LA items. Second, principal components analyses with
oblique rotation are applied to the satisfaction dimensions to see whether
the hypothesized dimensions were uni-dimensional, or a multi-facet struc-
ture is the cause of low item-to-total correlations in the previously men-
tioned dimensions.
In OHS and LA constructs, a single factor structure is obvious. On the
other hand, for LC, PS, and SR constructs the ratios of the first and second,
and the second and third eigenvalues are much higher than the ratio of any
of two other adjacent eigenvalues, indicating a distinct scree at two factors.
However, in two out of three cases the first two ratios are very close to each
other. Therefore, choosing the number of factors to extract becomes a dif-
ficult task. Since we do not want the decision to be purely based on the scree
criteria, which is known to be not very powerful and subjective (Zwick and
Velicer, 1986), we decided to apply Horn’s parallel procedure. Applying the
parallel analysis method with the procedure developed by Keeling (2000)
produced the parallel analysis criterion values shown in Table I, which also
includes the observed eigenvalues. Horn’s parallel analysis, which is the
most accurate method for selecting the appropriate number of factors,
suggests two underlying factors in both samples for all constructs except
one. In the SR scale, in the Kaleici sample Horn’s procedure proposes a
single factor solution. However, when the two-factor solution in Murat I
sample is investigated conceptually, we see that it separates the social133
COMPARISON OF URBAN HOUSING SATISFACTION
TABLE I
Observed and calculated eigenvalues for PLC, PS and SR constructs
Three item scales (PLC and PS)
Ten item scale (SR)
HPP Kaleici Murat I Kaleici Murat I HPP Kaleici Murat I
1.270
1.145
1.031 1.749
1.417
0.991 1.553
1.211
0.971 1.696
1.243
1.109 1.806
1.375
0.997 1.477
1.354
1.239 3.409
1.236
0.927 3.484
1.433
0.884
Note: HPP – Horn’s parallel procedure.
relationship satisfaction (SSR) from the perceived attitude toward the res-
ident (PAR). Since such a facet is commonsensical, we choose to extract two
factors from the SR dimension. Similarly LC and PS constructs are analyzed
conceptually and the following results are found to hold in both samples.
The two facets that emerge in the LC construct correspond to environmental
(ELC) and social (SLC) living conditions, while PS breaks down into sat-
isfaction with the house the subject lives in (SSH) and satisfaction from
physical characteristics of the neighborhood (SPC). The items that make up
these constructs are listed in the Appendix.
3.4.
Structural Model
Given the factor structure outlined above, we proceed with the analyses of
urban housing satisfaction in the two districts. First, the conceptual model
shown in Figure 1 is modified to the model shown in Figure 2.
In this new version of the model, the causal structure is kept constant, i.e.
the impact of all determinants of OHS flows through the perceived LC,
except minor modifications. The model is constructed in such a way that the
two facets of satisfaction with the PS influence perceived ELC, while the SR
satisfaction influences perceived SLC. Here one may also argue that a causal
link exists between perceived environmental living conditions and social
living conditions, flowing from SLC to ELC, such that the perception of
social living conditions can have an impact on the environmental living
conditions because if a resident is not satisfied with the atmosphere the
neighborhood offers, his/her perception of any environmental stimulus is
likely to be negatively influenced, or he/she may start to attend selectively to
the negative aspects of the surrounding environment. Or one may easily
argue that the reverse causality holds. In the empirical analysis, we134
NEVNIHAL ERDOGAN ET AL.
Fig. 2. Modified conceptual model. Note: OHS – overall housing satisfaction, ELC – envi-
ronmental living conditions, SLC – social living conditions, FQP – perceived quality of the
facilities, SSH – satisfaction with the house the subject lives in, SPC – satisfaction from physical
characteristics of the neighborhood, LA – satisfaction with the performance of the local
authorities, SSR – social relationship satisfaction, PAR – perceived attitude toward the resident.
estimated both models and found that such a causal link does not exist.
Therefore, in the following analyses, we let the two latent constructs be
correlated with each other but did not specify a causal link.
Since all factors in the conceptual model are latent variables, measured by
multiple items, a covariance structure model can be estimated to compare
the urban housing satisfaction in traditional and modern neighborhoods. 1
In doing that, we proceed in the following order. First, we estimated two
separate models for the two neighborhoods and check whether the proposed
relations hold. Secondly, we combined the two data sets and formally tested
the invariance of the basic structure of the constructs across two groups.
Last, we made quantitative comparisons of construct means across the two
groups. Analyses regarding the first step can be carried out by standard
covariance structure models, while the last two steps fall under the multi-
group analysis (measurement invariance) topic in covariance structure
models. Although there are a variety of techniques to assess measurement
invariance, we use the multiple-group confirmatory factor analysis model
proposed by Joreskog (1971) because it is accepted as the most powerful
approach. The main reason of choosing such an analysis agenda is as
follows. If evidence supporting the measures’ invariance is lacking, theCOMPARISON OF URBAN HOUSING SATISFACTION
135
conclusions drawn from the first step are to be considered as ambiguous or
erroneous, and comparisons of the latent construct means are meaningless,
since the measurement scales are fundamentally different across groups.
3.5.
Model Estimation
In this section we discuss the steps taken in the model estimation in detail.
The reader may jump to the findings section, without loss of substantial
information, if the steps followed in model building and the technical details
of the estimation procedure are not in his/her interest.
Initially, we ran the model shown in Figure 2 for Kaleici sample. A
thorough examination of the parameter estimates and modification indices
suggested two minor changes. The modification indices and the expected
parameter change statistics for two pairs of errors belonging to the OHS
construct are outstanding, suggesting a correlated errors model. Given that,
the decision for a correlated errors model should be supported by theoretical
arguments. We checked the wording of the four items to see whether these
item pairs had something more in common compared to the remaining items
in the construct. Item 13 ‘‘I feel a part of the neighborhood’’ and item 21
‘‘I feel at home here’’ have the concept of belongingness in common, while
item 29 ‘‘I think this neighborhood is very beautiful’’ and item 30 ‘‘This is a
wonderful place to live’’ share an underlying positive emotion. Note that
belongingness and positive emotions are not clearly underlying in any other
items of the OHS construct. Therefore, we allow these error terms to be
correlated.
Moreover, the factor loading of item 55 turned out to be insignificant.
Thus, we estimated a second model with the factor loading of item 55 set to
0. However, the model fit deteriorated dramatically. Although the factor
loading is insignificant, the high modification index and expected parameter
change for item 55 suggest that it should be estimated freely. Therefore, we
let the factor loading be freely estimated in the final, which gives a satis-
factory model fit. Although the v 2 is significant (v 2 (538) = 736.866,
p < 0.001), the RMSEA of 0.047 indicates an acceptable fit. The other most
commonly used practical fit indices are not above their recommended level
0.9 (CFI = 0.84, TLI = 0.82), but models with similar complexity 2 tend to
reveal such practical fit indices (Gerbing and Anderson, 1993). All factor
loadings are significant, and 26 out of 35 standardized factor loadings are
above 0.5.
The same correlated error structure emerges in the second sample as well.
However, this time three different items (43, 47, and 48) turn out to have136
NEVNIHAL ERDOGAN ET AL.
insignificant loadings. The fit statistics and the modification indices of the
model that exclude these items also suggest that the factor loadings should
be estimated freely although they are insignificant. The fit of the final model
is also reasonably satisfactory. As usual, v 2 is significant
(v 2 (473) = 693.145, p < 0.001), but practical fit indices indicate a reason-
able fit (RMSEA = 0.055, CFI = 0.82 and TLI = 0.80). All factor load-
ings in this model are significant, and 22 out of 33 standardized factor
loadings are above 0.5.
Next, we move on to the analyses of measurement invariance. Since the
purpose of the second step of the analyses is to make meaningful compar-
isons between the samples, we have to explore the basic meaning and
structure of the previously mentioned constructs in different neighborhoods.
We are specifically interested in whether the constructs can be conceptual-
ized in the same way across neighborhoods or not. Therefore, the following
multi-group measurement invariance tests were performed. First of all, the
scales should satisfy configural invariance condition, which is supported if
(1) the specified model with zero loadings on non-target factors fits data well
in all neighborhoods, (2) all salient factor loadings are significantly and
substantially different from zero, and (3) the correlations between the fac-
tors are significantly below unity, i.e. they are not redundant. In order to
make quantitative comparisons of construct means across the two groups,
the measures should exhibit metric and scalar invariance in addition to
configural invariance. Metric invariance provides for a stronger test of
invariance by introducing the concept of equal scale metrics across groups.
If an item satisfies this property, then the scores on the items making up a
construct can be meaningfully compared across the neighborhoods. On the
other hand, scalar invariance implies that cross-group differences in the
means of the observed items are due to the differences in the means of the
underlying constructs, and allows the researcher to compare the means of
the latent constructs across multiple groups.
Following the procedure explained above, we assess the measurement
invariance of the scales. The assessment has a sequential nature, and higher-
level models are nested in the lower-level models. We infer that the scales
exhibit measurement invariance if the fit of these nested models are not
deteriorated as we go up the ladder. For the comparison of these nested
models, v 2 difference test is commonly applied. However, v 2 difference test
suffers from the same problems as the v 2 -test. In large samples, virtually in
all cases, the null hypothesis is rejected. Therefore, some practical fit indices
are used to compare nested models, such as Consistent Akaike InformationCOMPARISON OF URBAN HOUSING SATISFACTION
137
TABLE II
Model comparisons for measurement invariance
Configural inv.
Partial metric inv.
Initial partial scalar inv.
Final partial scalar inv.
v 2 df RMSEA CAIC CFI TLI
1642.753
1663.911
1738.421
1712.476 1143
1166
1189
1188 0.0547
0.0536
0.0555
0.0540 3257.622
3122.357
3034.029
3015.041 0.801
0.801
0.781
0.791 0.780
0.785
0.768
0.778
Criteria (CAIC), Root Mean Square Error Approximation (RMSEA),
Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) (Table II).
We first estimate the configural invariance model. 3 It serves as the base-
line model against which other models are compared. The fit of the confi-
gural invariance model is satisfactory. Although the v 2 is significant
(v 2 (1143) = 1642.753, p < 0.001), the RMSEA of 0.055 indicates an
acceptable fit. As in the previous single model cases, the practical fit indices
are not above their recommended level 0.9 (CFI = 0.801, TLI = 0.780),
but this is due to model complexity. The CAIC for this model is 3257.622.
All, except the previously mentioned factor loadings, are significant in the
two neighborhoods. The 95% confidence intervals of correlation coefficients
between the latent constructs do not include the value 1. Therefore, we may
conclude that the urban housing satisfaction factors exhibited configural
invariance. Next, we test the hypothesis of metric invariance by constraining
the factor loadings of the common items to be invariant across neighbor-
hoods. The increase in v 2 is insignificant (Dv 2 (24) = 21.158, p < 0.629).
The fit does not decrease at all in terms of alternative fit indices. The
RMSEA of 0.054 indicated almost the same fit. CFI is 0.801, and TLI is
0.785. The CAIC for this model is 3122.357, indicating that the fit of the
model has actually improved. Thus partial metric invariance is also sup-
ported. 4
The final step is to impose scalar variance on the model. Given that only
partial metric invariance is achieved, intercepts of the items that satisfy
metric invariance condition are constrained to be equal across neighbor-
hoods. The increase in v 2 is significant (Dv 2 (24) = 74.51, p < 0.001).
Although the other fit indices do not show a dramatic deterioration in
model fit (RMSEA = 0.056, CFI = 0.781, TLI = 0.768, and CAIC =
3034.029), examination of the modification indices and the expected
parameter change statistics suggest that the intercept for a single item (S28,
MI = 31.359) is not invariant across neighborhoods. Relaxing these138
NEVNIHAL ERDOGAN ET AL.
constraints yields a modest and significant (Dv 2 (1) = 25.954, p < 0.001)
improvement in fit (RMSEA = 0.54, CFI = 0.791, TLI = 0.778, and
CAIC = 3015.041) compared to the initial partial scalar invariance.
Therefore, we may also conclude that the model satisfies partial scalar
invariance. The estimated factor loadings and item means with their asso-
ciated standard errors can be seen in Tables III and IV, respectively.
4.
FINDINGS
Having satisfied all necessary measurement invariance conditions, latent
construct means can now be compared safely. For the comparison of means,
we use Kaleici as the reference group and estimate the mean of the latent
constructs in the Murat I sample compared to the reference group. The
latent construct means and the corresponding standard errors are shown in
Figure 3, Panel (b). Means of three out of five exogenous constructs appear
to be significantly higher in Murat I. More specifically, PAR, satisfaction
with social relations (SSR) and satisfaction with the LA is higher in Murat I,
compared to Kaleici. Moreover, Murat I residents score significantly higher
than Kaleici residents in their satisfaction with the SLC as well as ELC.
However, in the modern neighborhood OHS is significantly lower than the
traditional neighborhood, Kaleici.
TABLE III
Factor loadings and item intercepts (s) for endogenous latent constructs (Kaleici/Murat I)
SLC
ELC
OHS
FQP
s
S2 1.000/–
3.137 (0.081)
S4 1.209 (0.223)
3.305 (0.085)
S44 0.812 (0.188)
3.682 (0.077)
S18
1.000/–
2.650 (0.087)
S43
1.054/0.316/(0.225)/(0.188)
2.613/3.167/(0.092)/(0.114)
S31
0.709 (0.139)
2.720 (0.079)
S13
1.000/–
3.730 (0.078)
S21
1.091 (0.091)
3.815 (0.080)
S22
1.040 (0.136)
3.127 (0.079)
S27
1.195 (0.142)
3.333 (0.083)
S28
1.160 (0.142)
3.425/3.976/(0.094)/(0.103)
S29
1.198 (0.142)
3.616 (0.082)
S30
1.217 (0.143)
3.517 (0.083)
S59
0.804 (0.128)
3.947 (0.074)
FQP
1.000/–0.917 (0.092)(0.076)
(0.077)
3.172
3.304
COMPARISON OF URBAN HOUSING SATISFACTION
139140
NEVNIHAL ERDOGAN ET AL.
Fig. 3. Latent construct means and structural coefficients. Note: Estimated path coefficients
(the numbers located on the lines), estimated latent construct means (the numbers located inside
the circles), and the associated standard errors (the numbers in parentheses). Bold numbers
indicate that the estimate is significantly different then zero.
These results could be related to the following reasons: Perceived attitude
towards the residents (PAR) and satisfaction with social relations (SSR) are
higher in Murat I since it has more homogenous socio-economic charac-
teristics. Satisfaction with the LA is also higher in Murat I because residentsCOMPARISON OF URBAN HOUSING SATISFACTION
141
of Murat I have better communication with local authorities due to their
educational level. SLC and ELC are significantly higher than conditions in
Kaleici due to its better conditions. However, the OHS is higher in Kaleici
than Murat I since the expectations of residents of Murat I are higher than
expectations of the people from Kaleici.
Next, we move on to discussion of the causal structure. For that, we first
restrict a set of coefficients shown in Figure 2 to be zero since they turn out
to be insignificant. The fit of the resulting model, shown in the two panels of
Figure 3, is satisfactory since the deterioration in the fit statistics is negli-
gible (v 2 (1199) = 1731.190, p < 0.001, RMSEA = 0.0556, CFI = 0.788,
TLI = 0.777, and CAIC = 2984.853). The following results emerge from
the analyses. In both samples, OHS is positively influenced by satisfaction
from social living conditions (SLC, 0.412, t-value = 2.736) and environ-
mental living conditions (ELC, 0.748, t-value = 4.494), but the drivers of
these latent constructs differ between traditional and modern neighbor-
hoods. In the traditional neighborhood case satisfaction from SLC is only
influenced by PAR, while satisfaction with the ELC is mainly driven by
satisfaction with (i) the physical characteristics of the neighborhood (SPC,
1.552, t-value = 2.978), (ii) performance of the local authorities (LA, 0.392,
t-value = 2.670), and (iii) perceived quality of the available facilities (FQP,
0.207, t-value = 3.928). In the modern neighborhood case, the former two
(SPC and LA) also influence satisfaction with the environmental living
conditions positively. Moreover, another latent construct, satisfaction with
the house that the subject lives in (SSH), is added to the set of drivers of
satisfaction with environmental living conditions (0.704, t-value = 2.409).
Surprisingly, in the Murat I sample we see that perceived quality of available
qualities (FQP) has no significant effect on satisfaction with environmental
living conditions. In the modern neighborhood, we find that satisfaction
with SLC construct is driven by both satisfaction with social relations (SSR,
0.240, t-value = 2.055) and perceived attitude toward the resident (PAR,
0.452, t-value = 2.356), as opposed to the traditional neighborhood,
Kaleici.
5.
SUMMARY AND CONCLUSION
This paper addresses some conceptual and measurement issues related to the
study of housing satisfaction. A thorough review of the existing literature on
housing satisfaction led us to a causal model of residential satisfaction
shown in Figure 1. In order to test the model we collected data from two
separate neighborhoods, which are characterized by historical and modern
backgrounds. For the data collection we modified the housing satisfaction142
NEVNIHAL ERDOGAN ET AL.
scale of Bardo and Dokmeci’s (1992). Using different tests we show that
housing satisfaction is indeed a multidimensional construct, consisting of
five dimensions, which are causally linked. These dimensions include overall
housing satisfaction, perceived living conditions, physical surroundings,
social relations, and local authorities.
As a result of our analyses we found that perceived attitude toward the
resident, satisfaction with social relations and satisfaction with the local
authority is higher in a modern neighborhood, compared to a traditional
neighborhood. We also found that residents of a modern neighborhood are
more satisfied with the social living conditions as well as the environmental
living conditions. Interestingly, in the traditional neighborhood, the overall
housing satisfaction is significantly higher compared to the modern neigh-
borhood.
An empirical examination of the model in traditional and modern
neighborhoods reveals that social and environmental living conditions
positively influence overall housing satisfaction. The results also indicate
that the drivers of the social and environmental living conditions constructs
differ between traditional and modern neighborhoods. We find that, in
traditional neighborhoods, social living conditions are influenced by per-
ceived attitude toward the resident, while environmental living conditions
are mainly driven by satisfaction from physical characteristics of the
neighborhood, satisfaction with the performance of the local authorities,
and perceived quality of the facilities. On the other hand, in modern
neighborhoods, perceived attitude toward the resident and social relation-
ship satisfaction turn out to be the main drivers of social living conditions.
As for the environmental living conditions, the influential constructs are
satisfaction from physical characteristics of the neighborhood, satisfaction
with the performance of the local authorities, and satisfaction with the
house the subject lives in.
5.1.
Contribution to Current Literature
Early empirical work on residential satisfaction used bivariate techniques to
identify the correlates of satisfaction within particular demographic groups
(e.g. African Americans, elderly) or in particular types of cities. More re-
cently, researchers have used multivariate techniques to test models of sat-
isfaction with three sets of variables: (1) individual demographic
characteristics and objective characteristics of the residential environment,
(2) intermediary variables consisting of assessments of different residential
characteristics, (3) residential satisfaction (Varady and Preiser, 1998). TheCOMPARISON OF URBAN HOUSING SATISFACTION
143
present study makes both academic and practical contributions. The aca-
demic contribution is to explore the nature of housing satisfaction, and then
develop a conceptual model of housing satisfaction and then to test the
model empirically. The practical contribution is to use the proposed mul-
tidimensional scale as a diagnostic tool to identify areas where specific
improvements are needed and to pinpoint aspects of housing satisfaction
that require work. Local authorities such as housing policy makers and
urban planners may use this framework to develop relevant and effective
strategies and to improve the dissatisfied conditions in the new and pe-
ripherial areas.
5.2.
Recommendations for Further Research
Researches should concentrate on developing neighborhood revitalization
strategies by improving home ownership, housing investment and economic
development. Also, the model and the scale of this research can be used by
other researchers for further research. This would also help to compare the
results of different areas with different studies.
APPENDIX: Questionnaire
OVERALL HOUSING SATISFACTION (OHS) 8 items – None Deleted Later
13 I feel a part of the neighborhood
21 I feel at home here
22 I find enough here to keep me busy
27 Life is not boring here
28 I wouldn’t prefer living in a different neighborhood. This place is not suitable for me
29 I think this neighborhood is very beautiful
30 This is a wonderful place to live
59 In general, I am quite happy with my life
PERCEIVED LIVING CONDITIONS (PLC) 11 items – 5 Items Deleted Later (*)
2 This neighborhood is very quite and neat
4 Families here do not let their children to disturb anybody
11 (*) Very few people here earn an adequate income
18 Nobody around here cares what the place look like
26 (*) Job opportunities here are as same as the other places
31 The green areas here make this place that can be lived in
31 The green areas here make this place that can be lived in
33 (*) The life standard here is influenced by national conditions
34 (*) The quality of life here is affected by national economic problems
41 (*) The social services provided here are as good as those in Edirne
43 The opportunities around here means that life here will continue to develop faster than
in other neighborhoods
44 Not much crime happens around here144
NEVNIHAL ERDOGAN ET AL.
PHYSICAL SURROUNDING (PS) 8 items – 2 Items Deleted Later (*)
23 My house meets my needs
24 This house is better than other houses I lived in before
25 Buildings here are beautiful as those in the place I lived before
36 (*) Houses here are just as good as those in Edirne
51 Most of the people in this neighborhood do not paint their houses on time
52 Most people in this neighborhood take care of their yards
55 The people who come from outside can find addresses easily
56 (*) Houses here are not too crowded together
SOCIAL RELATIONS (SR) 15 items – 5 Items Deleted Later (*)
1 It is very hard to find a real friend in this neighborhood
3 (*) Many people here believe that they act properly towards you
5 Everybody here is polite
7 Everybody here does not try to take advantage of you
8 When people live the mosque they forget the idea of brotherhood
10 If you are not the same as everyone else here you get mocked
14 People here are very stingy
15 (*) You have to spend a lot of money to be accepted around here
16 (*) Everyone here minds their own business
17 Everybody around here criticizes everybody else
32 I would like more neighbors nearby. Neighbors here live far away
49 (*) Owner of the houses in this neighborhood do not let their house to singles
50 (*) Old people are very well looked after in this neighborhood
57 Most people around here are interested in everybody’s personal business
58 There are not decent neighbor to form friendships with around here
LOCAL AUTHORITIES (LA) 5 items – None Deleted Later
12 The municipality here does whatever they want
20 The city provides very limited services
42 The city does not care about this neighborhood
47 Local officials do not listen our ideas
48 Nobody asks the people for their ideas
DIRECTLY EXCLUDED ITEMS – 12 items
6 The schools here prepare students well for the university
9 There is nobody in this community that can be a leader
19 I do not care what children do as long as they keep away from me
35 Compared to other neighborhoods, our shopping centers are wonderful
37 The health facilities here are as good as those in the other neighborhoods
38 Hospitals here provide a full range of services
39 The quality of life is same in old and new neighborhoods
40 Public buildings here are very well-kept
45 When my children grow up they will not find a house in this neighborhood to live in
46 There are not going to be enough jobs around here in the future
53 I like it here because it is close to my family
54 I like it here because good families live hereCOMPARISON OF URBAN HOUSING SATISFACTION
Photos from Kaleici neighborhood
145146
NEVNIHAL ERDOGAN ET AL.
Photo from Murat I neighborhood
NOTES
1
We performed all the analyses in LISREL 8.50.
Complexity refers both to sample size and model structure. Many of the relative fit indices are
affected by sample size, so that larger samples are seen as better fitting. Even though a fit index
may not include sample size in the formula it does not mean that the fit index is really inde-
pendent of the sample size. Therefore, we use a combination of the fit indices while reporting
model fit.
3
One necessary condition for multi-group measurement invariance tests is the equivalence of
number of items in different groups. Therefore during the measurement invariance tests items
43, 47, 48, and 55 are included in the set of items, and the associated parameters are estimated
freely. An alternative approach is to set the factor loadings of insignificant items to zero and the
associated error variances to infinitesimally small values in the relevant groups. When this
approach is followed, the fit of the model deteriorated compared to the freely estimated version,
and the outstanding modification indices suggested that the factor loadings should be freely
estimated. Note that this leads to partial measurement invariance, and partial invariance is still
a sufficient condition for meaningful comparisons.
4
For a meaningful comparison of latent construct means full measurement invariance is not a
necessary condition. Partial measurement invariance requires some of the estimated parameters
to be equal in multiple groups, and it is a sufficient condition for substantive analysis to be
meaningful. See Byrne et al. (1989) for further discussion.
2
REFERENCES
Altman, I. and C. Werner: 1985, Human Behavior and Environment: Advances in Theory and
Research, Vol. 8 – Home Environments (Plenum Press, New York).
Altman, I. and A. Wendersman: 1987, Human Behavior and Environment: Advances in
Theory and Research, Vol. 9 – Neighborhood and Community Environments (Plenum Press,
New York).
Bardo, J.W. and V Dokmeci: 1990, ÔA brief note on second-order community satisfaction
factors for two sub communities in Istanbul, TurkeyÕ, Multivariate Experimental Clinical
Research 9(3), pp. 145–149.COMPARISON OF URBAN HOUSING SATISFACTION
147
Bardo J.W. and V. Dokmeci: 1992, ‘Modernization, traditionalization and the changing
structure of community satisfaction in two-sub-communities in Istanbul, Turkey: A Pro-
crustean’, Genetic, Social and General Psychology Monographs, 118(3).
Barrasi, C., K.F. Ferraro and L.L. Hobey: 1984, ÔEnvironmental satisfaction, sociability and
well-being among the urban elderlyÕ, International Journal of Aging and Human Develop-
ment 18, pp. 277–293.
Bonaiuto, M. and A Aiello: 1999, ÔMultidimensional perception of residential environment
quality and neighborhood attachment in the urban environmentÕ, Journal of Environmental
Psychology 19, pp. 331–352.
Byrne, B.M., R.J. Shavelson and B. Muthen: 1989, ÔTesting for the equivalence of factor
covariance and mean structures: The issue of partial measurement invarianceÕ, Psychological
Bulletin 105(3), pp. 456–466.
DeVellis, R.F.: 1991, Scale Development: Theory and Applications (Sage, California).
Dokmeci, V. and L. Berköz: 2000, ÔResidential-location preferences according to demographic
characteristics in IstanbulÕ, Landscape and Urban Planning 48, pp. 45–55.
Duncan, T.: 1971, Measuring Housing Quality, Occasional Paper No. 20, Center for Urban and
Regional Studies, University of Birmingham.
Francescate, C., S. Weideman and J. Anderson: 1987, ÔResidential satisfactionÕ, in W. VlietVan
(ed.), Housing and Neighborhoods, Greenwood Press, London, pp. 43–57.
Fried, M. and P. Gleicher: 1961, ÔSome sources of residential satisfaction in an urban areaÕ,
Journal of the American Institute of Planners 27, pp. 305–315.
Gans, H.: 1967, ÔPlanning and city planning for mental healthÕ, in H. Eldredge (ed.), Taming
Metropolis (Praeger, New York).
Gerbing, D.W. and J.C. Anderson: 1993, ÔMonte Carlo evaluations of goodness of fit indices for
structural equation modelsÕ, in K.A. Bollen and J. Scott Long (eds.), Testing Structural
Equation Models (Sage, CA), pp. 40–65.
Goudy, W.J.: 1977, ÔEvaluation of local attributes and community satisfaction in small townsÕ,
Rural Sociology 42, pp. 371–382.
Goudy, W.J: 1982, ÔFurther consideration of indicators of community attachmentÕ, Social
Indicators Research 11, pp. 181–192.
Guest, A.M. and B.A. Lee: 1983, ÔConsensus on locality names within the MetropolisÕ, Soci-
ology and Social Research 67, pp. 374–391.
Hunter, A.: 1978, ÔPersistence of local sentiments in mass societyÕ, in D. Street (ed.), Handbook
of Contemporary Urban Life (Jossey-Bass, San Francisco, CA).
Hunter, A.: 1979, ÔThe urban neighborhood: Its analytical and social contextsÕ, Urban Affairs
Quarterly 14, pp. 267–288.
John, C. and F. Clark: 1984, ÔRace and social class differences in the characteristics desired in
residential neighborhoodsÕ, Social Science Quarterly 65, pp. 803–813.
Joreskog, K.G.: 1971, ÔSimultaneous factor analysis in several populationsÕ, Psychometrika 36,
pp. 409–426.
Kasarda, J.D. and M. Janowitz: 1974, ÔCommunity attachment in mass societyÕ, American
Sociological Review 39, pp. 328–339.
Keeling, K.B.: 2000, ÔA regression equation for determining the dimensionality of dataÕ, Mul-
tivariate Behavioral Research 35(4), pp. 457–468.
Marans, R.W. and W. Rodgers: 1975, ÔToward an understanding of community satisfactionÕ, in
A.H. Hawley and V.P. Rock (eds.), Metropolitan America in Contemporary Perspective,
Halstead, New York..
Merton, R.: 1968, Social Theory and Social Structure (Free Press, New York).
Michaelson, W.: 1977, Environmental Choice, Human Behavior and Residential Satisfaction
(Oxford University Press, New York).

You Need a Professional Writer To Work On Your Paper?