# RECORD 0f2b62072edfd32a439d1606541bbbb8547810e4
# TITLE: Willingness to pay for accessibility under the conditions of residential segregation
# YEAR: 2014.0  VENUE: nan  DOI: 10.3846/1648715x.2013.864342
# AUTHORS: Marko Kryvobokov; L. Bouzouina
# PAGES: 15  CANDIDATE TABLE PAGES (1-based): [6, 7, 8, 10, 14, 15]
# ABSTRACT: The hypothesis that different income groups have different willingness to pay for accessibility to the city centre is based on the standard monocentric model. This hypothesis is empirically tested with accessibility attributes in a hedonic model of apartment prices in the suburbs of the city of Lyon, France. The conditions of residential segregation are described, and apartment prices in the poor and the rich suburbs are analysed with regression techniques. Travel times to two urban centres are accounted for, as well as centrality and accessibility integral indexes. We found that in the selected areas the hypothesis is true. Spatial differences between the estimates for accessibility measures are significant. In more socially problematic areas, the willingness to pay for better accessibility is higher.


===== PAGE 1 (score 13, numeric-paren 1, table-words 0, var-words 12) =====

Copyright © 2014 Vilnius Gediminas Technical University (VGTU) Press Technika
http://www.tandfonline.com/TSPM
INTERNATIONAL JOURNAL OF STRATEGIC PROPERTY MANAGEMENT
ISSN 1648-715X print / ISSN 1648-9179 online
2014  Volume  18(2): 101–115
doi:10.3846/1648715X.2013.864342
* Corresponding author. E-mail: k_mark@ukr.net
or mark.kryvobokov@let.ish-lyon.cnrs.fr
WILLINGNESS TO PAY FOR ACCESSIBILITY UNDER 
THE CONDITIONS OF RESIDENTIAL SEGREGATION
Marko KRYVOBOKOV a,*, Louafi BOUZOUINA 
a Laboratory of Transport Economics (LET), 14, avenue Berthelot, F-69363 Lyon Cedex 07, France 
b Laboratory of Transport Economics (LET), ENTPE, 3, rue Maurice Audin, F-69518 Vaulx-en-Velin, France
Received 29 September 2012; accepted 16 January 2013
ABSTRACT. The hypothesis that different income groups have different willingness to pay for acces-
sibility to the city centre is based on the standard monocentric model. This hypothesis is empirically 
tested with accessibility attributes in a hedonic model of apartment prices in the suburbs of the city 
of Lyon, France. The conditions of residential segregation are described, and apartment prices in the 
poor and the rich suburbs are analysed with regression techniques. Travel times to two urban centres 
are accounted for, as well as centrality and accessibility integral indexes. We found that in the selected 
areas the hypothesis is true. Spatial differences between the estimates for accessibility measures are 
significant. In more socially problematic areas, the willingness to pay for better accessibility is higher.
KEYWORDS: Accessibility; Bid-rent function; Hedonic regression; Apartment price; Residential seg-
regation
REFERENCE to this paper should be made as follows: Kryvobokov, M.; Bouzouina, L. 2014. Willing-
ness to pay for accessibility under the conditions of residential segregation, International Journal of 
Strategic Property Management 18(2): 101–115.
1. INTRODUCTION
Economic literature about urban segregation ex-
perienced rapid development since the first work 
of Kain (1968) on the difficulties of employment 
access for the inhabitants of American ghettos. 
Numerous case studies dealing with the question 
of neighbourhood effect test the impact of social 
polarisation or accessibility on unemployment, iso-
lation, crime or school results in problematic dis-
tricts of different cities and countries (Case, Katz 
1991; Hoxby 2000; Bayer et al. 2008; Galster et al. 
2008). 
In France, the economic studies of segregation 
are more recent and mainly dated by the years 
2000s (Fitoussi, Savidan 2003; Maurin 2004; 
Buisson, Mignot 2005). With several exceptions 
analysing geographic and ethnic origins (Pan Ké 
Shon 2010), this literature focuses on the socio-
economic dimensions to demonstrate the increas-
ing inequalities between rich and poor neighbour-
hoods and to measure their negative consequences 
on the economic, human and social capital of poor 
inhabitants, whose significant share is suffered 
from high unemployment (Bouabdallah et al. 2002; 
Choffel, Delattre 2003; Gaschet, Gaussier 2005; 
Gobillon, Selod 2004; Korsu, Wenglenski 2010). 
Most of these studies confirm the hypothesis that 
the unemployment of poor residents is partially 
caused by poor accessibility to jobs (especially to 
low-skilled jobs) and high poverty rates. The two 
main reasons of poor employment accessibility in 
the Paris region are location in remote neighbour-
hoods with few jobs and low car-ownership rate 
(Korsu, Wenglenski 2010). In their analysis of 
the Lyon, Marseille, and Lille urban agglomera-
tions, Mignot et al. (2009) highlight the point that 
though some recent studies report little inequal-
ity in daily mobility since households have access 
to cars, at the same time inequality increases for 
those without car. 
The growing interest of academic research to 
residential segregation in France can be explained 
by a complicated reality including violence and 
b


===== PAGE 6 (score 35, numeric-paren 0, table-words 7, var-words 19) =====

106
M. Kryvobokov, L. Bouzouina
The described four suburbs are the examples of 
the uneven distribution of social housing stock. In 
the western suburbs, this percentage is lower than 
0.20, especially in Sainte-Foy-lès-Lyons, while in 
the eastern banlieues half of population lives in 
social housing (Table 1). 
Income disparities and social contrasts between 
the selected suburbs clearly demonstrate that in 
terms of attractiveness and prestige, “bourgeois” 
western areas are far surpassing their eastern 
counterparts. Moreover, both Vaulx-en-Velin and 
Vénissieux have an image of well-known high-
poverty banlieues. The following quotation from 
Korsu and Wenglenski (2010) about the public 
view of problematic areas in general can be ex-
ploited to describe both: “many people associate 
such neighbourhoods with violence, criminality, 
unwillingness to work, alcohol and drug addiction, 
etc.” (ibid: 2285).
In respect to public transport, the socialist gov-
ernance of the Lyon agglomeration tries to favour 
eastern suburbs. The most expensive investments 
in public transport, such as the extensions of met-
ro and tramway lines, take place to the east from 
Lyon, including Vénissieux and Vaulx-en-Velin, 
while in the western suburbs these types of pub-
lic transports do not exist. In the eyes of many 
Lyonnais, the relative underdevelopment of public 
transport in western suburbs is to same degree 
a deliberate policy supporting their higher stand-
ing. A good illustration of the Not-In-My-BackYard 
principle of the wealthy population is their protest 
(though unsuccessful) in the 1990s against prolon-
gation of the metro line directly linking Vénissieux 
with north-western Lyon. 
We use the data on about 10,000 apartment 
sales selected randomly from all sales in the cen-
tral part of the Lyon metropolitan area in the 
period 1997–2008. These data on sale prices and 
apartment attributes were provided by Perval, who 
collects information about real estate transactions 
in France. After deleting observations with incom-
plete data, 4,362 apartments remained. The defini-
tion of variables and descriptive statistics are given 
in Table 2. It contains information about transac-
tions, as well as on apartment attributes and loca-
tion attributes. It does not include the number of 
rooms, because the cardinal discrete variables for 
them are highly correlated with apartment area 
and are not significant if included in the hedonic 
model. Many observations contain no data about 
the number of parking places, number of cellars, 
and quality of view; therefore the specific dummy 
variables are created. In the sample, 28% of the 
apartments were new, sold in a primary market. 
The location variables in Table 2 include the 
percentages of high-, middle-, and low-income 
households, and dummies for location in one of 
four ad hoc districts. The four ad hoc districts are 
quite large, but relatively homogenous territories, 
divided by water frontiers and the boundaries of 
the urbanised area. District 1 includes the most 
central part of Lyon between the Rhône and the 
Saône. District 2 is an urbanised area on the left 
bank of the Rhône. District 3 is an urbanised area 
on the right bank of the Saône. District 4 is the 
less urbanised territory. 
Table 2 contains the following accessibility 
measures: travel times by car, travel times by pub-
lic transport, tertiary employment centrality index, 
and tertiary employment accessibility index. The 
origin-destination (O-D) matrix of travel times is 
estimated for morning peak with the MOSART 
transportation model developed with the VISUM 
software (Bonnafous et al. 2011). It is a four-step 
model, which forecasts transportation demand. We 
account for the travel times to the Lyon City Hall 
and to Bellecour, which is the main public transport 
junction located in the central part of Lyon. Both 
City Hall and Bellecour are among the fifteen (sub)
centres identified with a regression of service em-
ployment density on travel time (Kryvobokov 2010), 
Table 1. Socio-demographic description of the four suburbs
Attribute
Sainte-Foy-lès-Lyons
Oullins
Vénissieux
Vaulx-en-Velin
Population
22,616
26,426
57,566
40,241
Households
9,208
12,032
21,965
14,293
Persons in household
2.46
2.20
2.62
2.82
Cars to households ratio
1.25
1.01
0.95
0.87
Cars to population ratio
0.51
0.45
0.37
0.31
Percentage of high-income households
0.17
0.14
0.11
0.10
Percentage of low-income households
0.21
0.25
0.33
0.37
Percentage unemployed
0.07
0.09
0.18
0.21
Percentage of households in social housing
0.07
0.15
0.50
0.51


===== PAGE 7 (score 39, numeric-paren 0, table-words 4, var-words 22) =====

107
Willingness to pay for accessibility under the conditions of residential segregation
similarly to McDonald and Prather (1994). The cen-
trality and accessibility indexes are integral meas-
ures, which take account of tertiary employment. 
The former is gravity-based and the latter is a “sub-
jective” measure based on satisfaction thresholds 
and fuzzy logic, for details see Appendix 1. Note 
that in Table 2 the geographical unit of analysis is 
not municipality, but smaller zone; that is why the 
percentages of income groups are lower and higher 
in comparison with Table 1, which represents mu-
nicipalities. The zones are French statistical units 
IRISes (Îlots Regroupés pour l’Information Statis-
tique) close to the tracts in the US.
Of this sample, we select four sub-samples from 
the analysed municipalities (Table 3). The average 
floor area is the highest in Sainte-Foy-lès-Lyons. 
Though this attribute is the lowest in Oullins, it 
becomes the second highest there if calculated 
per person using the average number of persons 
in household in municipality. According to the 
average price per square meter, Sainte-Foy-lès-
Lyons is again the leader. The minimum prices 
in the western suburbs are much higher than in 
the eastern ones, but relatively high average price 
and very high maximum in Vénissieux indicates 
a wide diversity of its housing stock. According to 
the t-test, Vaulx-en-Velin and Vénissieux, unlike 
western suburbs, have equal means for some at-
tributes, e.g. the percentage of low-income house-
holds.
Table 2. Descriptive statistics of apartment attributes
Variable
Description
Mean
Minimum
Maximum
Std. deviation
Price
Transaction price, Euro
122,507.50
10,965
1,120,000
77,200
Year97-Year08
Dummies for year of transaction
0.04-0.13
0
1
0.20-0.34
Area
Apartment area, square metres
68.94
8
301
28.12
Bath1-Bath3
Dummies for number of bathrooms
<0.01–0.93
0
1
0.05–0.25
ParkUn
Dummy for cases with no data  
about parking places
0.26
0
1
0.44
Park0-Park3
Dummies for number of parking places
<0.01–0.49
0
1
0.06–0.50
FloorGr
Dummy for ground floor
0.13
0
1
0.33
Floor1
Dummy for storey 1
0.18
0
1
0.39
Floor2_4
Dummy for storey 2 to 4
0.49
0
1
0.50
Floor5_8
Dummy for storey 5 to 8
0.18
0
1
0.38
Floor9+
Dummy for storey 9 or more
0.02
0
1
0.14
Constr<1850-
Constr1992<
Dummies for period of construction
0.03–0.34
0
1
0.17–0.47
CondGood
Dummy for good state
0.81
0
1
0.39
CondMed
Dummy for state when some  
maintenance is needed
0.16
0
1
0.37
CondBad
Dummy for state when renovation  
is needed
0.03
0
1
0.17
ViewNo
Dummy for cases with no data  
about view
0.60
0
1
0.49
ViewGood
Dummy for view increasing value
0.38
0
1
0.48
ViewBad
Dummy for view decreasing value
0.02
0
1
0.13
Cellar0-Cellar2
Dummies for number of cellars
0.02–0.66
0
1
0.13–0.47
Garden
Dummy for existence of garden
0.05
0
1
0.22
Terrace
Dummy for existence of terrace
0.09
0
1
0.29
NewApartment
Dummy for new apartment
0.28
0
1
0.45
%HighIncome
Percentage of high-income households
12.55
4.34
28.77
2.91
%MidIncome
Percentage of middle-income households
57.96
42.70
66.20
3.30
%LowIncome
Percentage of low-income households
29.49
10.24
52.12
5.78
District1-District4
Dummies for location in districts
0.01–0.64
0
1
0.12–0.48
TTBellecour, 
TTCityHall 
Travel time to Bellecour, travel time  
to City Hall in minutes
11.00–21.73
0.55
29.32–51.02
4.99–11.16
Centrality Index
Centrality index
37.04–42.13
12.15–13.45
100.00
15.68–17.23
Accessibility Index
Accessibility index
63.07–64.70
0.34–0.65
100.00
25.63–27.47


===== PAGE 8 (score 26, numeric-paren 0, table-words 6, var-words 11) =====

108
M. Kryvobokov, L. Bouzouina
Table 3 gives the average values of accessibility 
measures in the four municipalities. The eastern 
suburbs are located in general a bit farther from 
the two urban centres, but their indexes for public 
transport are higher at the expense of better devel-
oped network. The number of zones in Sainte-Foy-
lès-Lyons, Oullins, Vaulx-en-Velin and Vénissieux 
are 10, 11, 18 and 24 respectively (see Fig. 1).
Besides using the sample of 4,362 observations, 
we extract the 20% most expensive apartments as 
well as the 20% cheapest apartments assuming 
that their buyers belong rather to the rich and 
the poor households respectively. These sub-sam-
ples called hereafter “expensive” and “cheap” are 
used in further analyses as proxies for the non-
geographical submarkets for the rich and the poor, 
because it is difficult to find information about 
the income of the buyers of particular apartments 
in quantity sufficient to create a mass valuation 
model. For these two sub-samples the equality of 
means hypothesis is rejected for price, floor area, 
income levels and all accessibility measures. We 
admit that this division does not correspond to 
the rich and the poor household groups described 
above. 
Furthermore, this sampling can be biased as 
some share of expensive apartments could be ex-
pensive just because of their large floor area, but 
cheap in respect to area unit. Some observations in 
the cheap segment, on the contrary, can be cheap 
due to small floor area, but have high price per 
area unit. In the context of the study, however, 
it is logical to assume that the poor are not able 
to buy expensive apartments even if their prices 
per area unit are low; while the rich prefer rather 
bigger accommodations. Therefore we suppose that 
our sampling strategy is less biased than it could 
be if it was based on price per square meter. 
Correlation of apartment price per square me-
ter (deflated to the initial year 1997) with acces-
sibility measures are in line with our hypothesis: 
prices of more expensive apartments are less cor-
related with accessibility. If to consider the sales 
in Sainte-Foy-lès-Lyons and Vénissieux, their cor-
relations behave similarly to the expensive and 
cheap sub-samples respectively. In Oullins and 
Vaulx-en-Velin, the sign of correlation coefficient 
is unstable. To understand the nature of these 
anomalies, econometric analysis is needed. 
4.2. Hedonic regression
Applying the hedonic regression model of price as 
the dependent variable with willingness to pay for 
different attributes as independent variables, we 
continue the tradition proposed by Rosen (1974). If 
the supply of each characteristic is perfectly elas-
tic, hedonic coefficients reveal demand for charac-
teristics, though in most real-world contexts such 
a stringent maintained hypothesis is untenable 
(Malpezzi 2003). Hedonic price model is widely 
used in real estate research. The recent examples 
of its application to housing prices in France are 
Table 3. Apartment sales and accessibility in the four suburbs
Attribute
Sainte-Foy-lès-Lyons
Oullins
Vénissieux
Vaulx-en-Velin
Apartment sales
Number of sales
84
100
97
54
Average floor area, m2
79
63
67
65
Average floor area per average 
number of persons, m2
32.1
28.6
25.6
22.9
Average price per m2, Euro
1,611
1,495
1,464
1,171
Minimum price per m2, Euro
675
393
226
307
Maximum price per m2, Euro
3,548
3,442
3,674
2,631
Accessibility by car
TTBellecour, minutes
12.17
10.62
14.45
21.13
TTCityHall, minutes
14.67
14.46
17.61
17.93
Centrality Index
23.82
25.61
26.02
23.73
Accessibility Index
28.20
33.94
30.77
23.10
Accessibility by public transport
TTBellecour, minutes
28.61
29.42
29.73
35.41
TTCityHall, minutes
31.48
33.33
35.08
32.69
Centrality Index
20.84
20.03
23.13
21.74
Accessibility Index
31.65
27.44
38.72
35.48


===== PAGE 9 (score 33, numeric-paren 0, table-words 9, var-words 15) =====

109
Willingness to pay for accessibility under the conditions of residential segregation
Gouriéroux and Lafarrère (2009) and Baltagi and 
Bresson (2011).
There is no strong theoretical basis for substan-
tiating the correct functional form of a hedonic re-
gression (Halvorsen, Pollakowski 1981; Malpezzi 
2003), but empirical tests found that the log-linear 
form has a number of advantages over the linear one 
(see e.g. Follain, Malpezzi 1980). More general and 
flexible form was proposed by Box and Cox (1964), 
but this function is beyond our current study. 
In our specification, the dependent variable is 
the logarithm of Price. Among the independent 
variables described in Table 2, the logarithmic 
transformations of Area, %MidIncome, %High-
Income and travel times are used. The following 
attributes are the default values: Year97, Bath1, 
Park0, FloorGr, Constr1981_1991, CondGood, 
ViewGood, Cellar0 and District1. 
First, an overall Ordinary Least Squares (OLS) 
model is built with 4,362 observations. Separate 
OLS models are created for the expensive and 
cheap apartment price segments; according to the 
t-test of equality of means of regressors, the null 
hypothesis is rejected for the majority of variables, 
including accessibility attributes and apartment 
floor area.
The observations from each of the four selected 
suburbs are exploited with separate OLS and Spa-
tial Error models. In the latter, the error term is 
a function of the residuals in neighbouring areas. 
The geographically weighted regression method-
ology (GWR) is applied to all available sales to 
solve a regression equation in each observation 
point. For the OLS, a weighted standard error 
test (Schnare, Struyk 1976) shows that the four 
submarkets significantly reduce standard error 
in comparison with the pool model. The t-tests of 
equality of regression coefficients in the four sub-
markets give the following results: for accessibility 
measures, the null hypothesis is rejected in most 
cases, while for some other attributes, e.g. floor 
area, the coefficients are rather stable. 
Each model is run eight times to examine the 
willingness to pay for each of the accessibility at-
tributes in question, i.e. travel times to two urban 
centres and centrality and accessibility indexes, 
while each attribute is measured twice: for car and 
for public transport. Our hypothesis is that the 
richer population evaluates travel time to city cen-
tres differently from the poor. We examine as well 
if this is true in respect to the integral indexes. 
Under the conditions mentioned in Section 2, the 
absolute value of the estimate for the accessibility 
in the hedonic price model should be lower for the 
rich than for the poor. 
Table 4 represents the estimated hedonic coeffi-
cients as well as model fit for OLS and GWR mod-
els. For the OLS models, the maximum variance 
inflationary factor (VIF) is also reported. All the 
significant coefficients of the models with TTBelle-
cour by car are presented in Appendix 2, where the 
estimates of the expensive and cheap sub-samples 
and median GWR results are reported as well. 
The three OLS models support our hypothesis. 
In the expensive segment, the absolute values of 
the estimates are always lower (often more than 
twice) than those in the overall model. In the cheap 
segment, the absolute values of the accessibility 
coefficients are always higher than those in the 
expensive segment. The negative effect of increase 
in travel time to Bellecour by public transport by 
1% leads to price reduction in the cheap segment 
by 0.17%, which is 2.8 times more than the price 
reduction in the expensive segment. 
The four municipalities can also be analysed 
as submarkets, for which individual models are 
created. The same variables except construction 
years and some structural dummies are used in 
the OLS and Spatial Error (with a threshold dis-
tance weighting scheme) models for the submar-
kets. The Lagrange multiplier (LM) tests, applied 
for the OLS models, indicate high significance level 
of spatial errors in each of the four suburbs. The 
LM tests of the spatial lags are significant in Oul-
lins and Vaulx-en-Velin, but due to difficulty in 
estimates interpretation, Spatial Lag model is not 
applied in this study. The estimates for the ac-
cessibility measures are presented in Appendix 3. 
However, this approach is rather useful to study 
the willingness to pay for accessibility measures 
within submarkets than the differences between 
them. Indeed, the estimates are insignificant and 
have “wrong” signs in Oullins and Vaulx-en-Velin, 
where the observations are less dispersed spatially 
than in Sainte-Foy-lès-Lyons and Vénissieux. The 
metro station in Vaulx-en-Velin opened in Octo-
ber 2007, by the end of the study period, and this 
fact can contribute to the “wrong” signs of acces-
sibility variables. The effect of the new station on 
the price dynamics is however beyond the current 
study limits. In Sainte-Foy-lès-Lyons, the coeffi-
cients have “right” signs, but they are insignifi-
cant in most cases. The most intuitive results in 
Vénissieux with its dispersed observations are ob-
tained with the OLS technique, but when spatial 
effect is accounted for, the accessibility measures 
calculated for car lose significance. Thus, within 
Vénissieux public transport has more impact on 
housing prices. 


===== PAGE 10 (score 85, numeric-paren 41, table-words 13, var-words 4) =====

110
M. Kryvobokov, L. Bouzouina
Concerning the geographically weighted results 
in Table 4, the average estimates among all the 
observations are reported as well as the four ex-
tractions from the examined suburbs taken from 
the overall GWR models. The absolute values of 
the average GWR estimates, which are a bit lower 
than the OLS results, increase from municipality 
to municipality in the sequence presented in Table 
4. The only exception is that for the accessibility 
index for car in Oullins, whose coefficient is signifi-
cant for only 56% of observations. In the eastern 
suburbs, the percentage of significant cases is al-
ways higher than 0.90. Vaulx-en-Velin, where this 
percentage is always equal to unity, demonstrates 
the biggest difference from Sainte-Foy-lès-Lyons. 
The ratio of their coefficients is 5.7 for the travel 
time to Bellecour by car. The absolute values of the 
estimates for the western suburbs are, as a rule, 
lower than the GWR average, though Oullins is 
an exception in the three cases of public transport 
(travel times to Bellecour, City Hall and centrality 
index), which can be explained by relative under-
development of public transport at the time, for 
which the data have been collected. In the eastern 
suburbs, the absolute values of the accessibility es-
timates are almost always higher than the overall 
GWR averages. 
5. CONCLUSIONS
The monocentric-based hypothesis that willingness 
to pay for better accessibility for the poor is higher 
than for the rich is empirically supported with a 
hedonic model of apartment prices. Monocentric 
models with travel times to two versions of city 
centre are supplemented with integral centrality 
and accessibility indexes. The results of hedonic 
models with different accessibility measures are 
quite similar. First, the hypothesis is supported 
with the cheap and expensive apartment segments. 
Table 4. Estimates for the accessibility measures 
Model (OLS),  
extraction (GWR)
Accessibility by car
Accessibility by public transport
TTBellecour
TTCityHall
Centrality 
Index
Accessibility 
Index
TTBellecour
TTCityHall
Centrality 
Index
Accessibility 
Index
OLS*
Overall,
4,362 obs., 
adj. R2 = 0.84 – 0.85, 
Max VIF = 6.27
–0.224
(<0.001)
–0.278
(<0.001)
0.0085
(<0.001)
0.0051
(<0.001)
–0.223
(<0.001)
–0.228
(<0.001)
0.0078
(<0.001)
0.0049
(<0.001)
Expensive,
873 obs., 
adj. R2 = 0.84 – 0.86, 
Max VIF = 8.02
–0.093
(<0.001)
–0.129
(<0.001)
0.0042
(<0.001)
0.0024
(<0.001)
–0.062
(<0.001)
–0.083
(<0.001)
0.0028
(<0.001)
0.0011
(0.017)
Cheap,
872 obs., 
adj. R2 = 0.70–0.72, 
Max VIF = 9.18
–0.176
(<0.001)
–0.217
(<0.001)
0.0069
(<0.001)
0.0043
(<0.001)
–0.173
(<0.001)
–0.157
(<0.001)
0.0060
(<0.001)
0.0038
(<0.001)
GWR**
Overall average, 
4,362 obs.,
R2 = 0.87 – 0.88
–0.175
(0.96)
–0.235
(0.97)
0.0071
(0.96)
0.0049
(0.95)
–0.151
(0.93)
–0.159
(0.97)
0.0052
(0.97)
0.0040
(0.91)
Sainte-Foy-lès-Lyons,
84 obs.
–0.106
(0.95)
–0.169
(1.00)
0.0061
(1.00)
0.0025
(0.87)
–0.104
(1.00)
–0.127
(1.00)
0.0044
(1.00)
0.0022
(0.99)
Oullins,
100 obs.
–0.126
(0.73)
–0.173
(0.83)
0.0066
(0.99)
0.0018
(0.56)
–0.202
(1.00)
0.192
(1.00)
0.0088
(1.00)
0.0033
(1.00)
Vénissieux,
97 obs.
–0.169
(0.90)
–0.307
(0.93)
0.0105
(1.00)
0.0049
(1.00)
–0.322
(1.00)
–0.306
(1.00)
0.0125
(1.00)
0.0060
(1.00)
Vaulx-en-Velin,
54 obs.
–0.599
(1.00)
–0.505
(1.00)
0.0133
(1.00)
0.0070
(1.00)
–0.520
(1.00)
–0.398
(1.00)
0.0141
(1.00)
0.0080
(1.00)
Notes: * – significance level is in the parentheses; ** – percentage of cases significant at the 5% level is in the paren-
theses.


===== PAGE 11 (score 15, numeric-paren 0, table-words 1, var-words 13) =====

111
Willingness to pay for accessibility under the conditions of residential segregation
Second, the support is shown with the example of 
the suburbs representing the patterns of residen-
tial segregation: the inhabitants of poor suburbs 
are willing to pay several times more for the acces-
sibility components of apartment price. Moreover, 
spatial differences in perception of accessibility are 
much higher than those between the expensive 
and cheap non-geographical apartment segments.
The spatial dispersion of observations in the 
examined four sub-samples is important, when 
the attributes within sub-sample are examined. 
Within one poor suburb, accessibility measures are 
found significant for public transport, but not for 
car. The GWR methodology allows comparing the 
effect between sub-samples in the overall model. 
In the selected eastern Lyon suburbs we have 
not found, in general, poorer accessibility, espe-
cially when public transport is taken into account. 
What is really different is the willingness to pay 
for accessibility: the more socially problematic is 
the area, the higher is the price for better acces-
sibility. This can be connected with lower car-own-
ership rate and other problems of disadvantaged 
neighbourhoods. From this viewpoint, the prior-
ity of public transport development in the east-
ern suburbs of Lyon looks reasonable as here its 
positive impact is more significantly capitalised in 
housing prices. Thus, investment in transport in-
frastructure in poor areas, which decreases travel 
time and simultaneously increases housing value, 
can be regarded in terms of economic efficiency 
and not only in those of social justice. 
The standard theoretical model presented in 
the paper does not include location amenities 
other than the proximity to the CBD. Some em-
pirical extension is possible with integral meas-
ures of centrality and accessibility considering 
service employment. In future study, more atten-
tion should be paid to other location amenities and 
their place in the trade-off with accessibility in the 
utility functions of household location choice of so-
cial groups. The techniques of Location Value Re-
sponse Surface and Market Basket Value (McClus-
key et al. 2000; Borst, McCluskey 2007; D’Amato 
2010) can be potentially useful to identify value 
influence centres and to be applied as a stage of 
market segmentation.
ACKNOWLEDGEMENTS
This study is part of project PLAINSUDD spon-
sored through French ANR (number ANR-08-
VD-00). The authors are grateful to Perval and 
Pierre-Yves Péguy for providing data on apart-
ment prices and attributes and to Nicolas Ovtracht 
for calculating the coordinates of apartment blocks 
and travel times with MOSART. The paper has 
been benefited from the discussions with Alain Bon-
nafous, Yves Crozet and Charles Raux as well as 
from valuable comments of anonymous reviewers.
REFERENCES
Alonso, W. 1964. Location and land use. Cambridge, 
MA: Harvard University Press.
Anas, A.; Arnott, R.; Small, K. A. 1998. Urban spatial 
structure, Journal of Economic Literature 36(3): 
1426–1464.
Baltagi, B. H.; Bresson, G. 2011. Maximum likelihood 
estimation and Lagrange multiplier tests for panel 
seemingly unrelated regressions with spatial lag 
and spatial errors: an application to hedonic housing 
prices in Paris, Journal of Urban Economics 69(1): 
24–42. http://dx.doi.org/10.1016/j.jue.2010.08.007 
Bayer, P.; Ross, S. L.; Topa, G. 2008. Place of work and 
place of residence: informal hiring networks and la-
bor market outcomes, Journal of Political Economy 
116(6): 1150–1196. http://dx.doi.org/10.1086/595975 
Bonnafous,  A.; Crozet,  Y.; Mercier,  A.; Ovtracht,  N.; 
Péguy, P. -Y.; Puech, F. 2011. MOSART et le projet 
PLAINSUDD: une plate-forme de modélisation et 
simulation de l’accessibilité pour l’aide à la décision 
et l’aménagement du territoire, in J.-P. Antoni (Ed.). 
Modéliser la ville: formes urbaines et politiques de 
transport. Paris: Economica, 186–210. 
Borst, R.; McCluskey, W. 2007. Comparative evaluation 
of the comparable sales method with geostatistical 
valuation models, Pacific Rim Property Research 
Journal 13(1): 106–130.
Bouabdallah, K.; Cavaco, S.; Lesueur, J.-Y. 2002. Re-
cherche d’emploi, contraintes spatiales et durée de 
chômage: une analyse microéconométrique, Revue 
d’Economie Politique 1: 137–157.
Bouzouina, L. 2008. Ségrégation spatial et dynamique 
métropolitaines. Thèse de Doctorat en Sciences Eco-
nomique. Université Lumière Lyon 2.
Bouzouina, L. 2009 Densité et ségrégation socio-spatiale 
au sein des villes françaises, in F. Gaschet, C. La-
cour (Eds.). Métropolisation et ségrégation. Bordeaux: 
Presses Universitaires de Bordeaux, 120–134,
Box, G. E. P.; Cox, D. 1964. An analysis of transforma-
tions, Journal of the Royal Statistical Society, Series 
B-Statistical Methodology 26(2): 211–252. 
Buisson, M.-A.; Mignot, D. 2005 Concentration écono-
mique et ségrégation spatiale. Bruxelles: De Boeck.
Case, A. C.; Katz, L. F. 1991 The company you keep: the 
effect of family and neighborhood on disadvantaged 
youths, National Bureau Economic Research Work-
ing Papers, 3705.
Choffel, P.; Delattre, E. 2003. Habiter un quartier défa-
vorisé: quels effets sur la durée de chômage?, Pre-
mières synthèses 43.1, Dares.
Goffette-Nagot, F.; Thomas, I.; Zenou, Y. 2000. Struc-
ture urbaine et revenus des ménages, in C. Baumont, 
P. Combes, P.-H. Derycke; H. Jayet (Eds.). Economie 
géographique: approches théoriques et empiriques. 
Paris: Economica, 276–302.


===== PAGE 14 (score 97, numeric-paren 58, table-words 2, var-words 7) =====

114
M. Kryvobokov, L. Bouzouina
APPENDIX 2 
OLS and GWR models with TTBellecour by car
Variable,
parameter
OLS
GWR
Overall
Expensive
Cheap
Constant
4.552 (<0.001)
7.186 (<0.001)
6.947 (<0.001)
4.255
TTBellecour
–0.224 (<0.001)
–0.093 (<0.001)
–0.176 (<0.001)
–0.136
Year99
0.092 (0.001)
0.090 (0.02)
–0.087 (0.26)
0.062
Year00
0.162 (<0.001)
0.157 (<0.001)
–0.029 (0.71)
0.136
Year01
0.249 (<0.001)
0.281 (<0.001)
0.204 (0.01)
0.220
Year02
0.310 (<0.001)
0.330 (<0.001)
0.223 (0.004)
0.279
Year03
0.474 (<0.001)
0.485 (<0.001)
0.402 (<0.001)
0.451
Year04
0.633 (<0.001)
0.589 (<0.001)
0.590 (<0.001)
0.611
Year05
0.801 (<0.001)
0.748 (<0.001)
0.749 (<0.001)
0.764
Year06
0.933 (<0.001)
.801 (<0.001)
0.912 (<0.001)
0.897
Year07
1.011 (<0.001)
0.900 (<0.001)
1.055 (<0.001)
0.983
Year08
1.003 (<0.001)
0.916 (<0.001)
1.010 (<0.001)
0.963
Area
0.932 (<0.001)
0.690 (<0.001)
0.383 (<0.001)
0.937
Bath2
0.094 (<0.001)
0.046 (0.001)
0.053 (0.78)
0.081
ParkUn
0.092 (<0.001)
0.016 (0.48)
0.043 (0.09)
0.043
Park1
0.154 (<0.001)
0.039 (0.20)
0.090 (0.001)
0.099
Park2
0.193 (<0.001)
0.080 (0.01)
0.067 (0.58)
0.139
Floor1
0.067 (<0.001)
0.027 (0.36)
0.078 (0.01)
0.072
Floor2_4
0.104 (<0.001)
0.038 (0.16)
0.103 (<0.001)
0.118
Floor5_8
0.133 (<0.001)
0.073 (0.01)
0.134 (<0.001)
0.147
Floor9+
0.132 (<0.001)
0.166 (0.002)
0.194 (0.01)
0.165
Constr1850_1913
–0.092 (<0.001)
0.042 (0.18)
–0.119 (0.01)
–0.083
Constr1914_1947
–0.082 (<0.001)
0.112 (0.001)
–0.110 (0.01)
–0.062
Constr1948_1969
–0.153 (<0.001)
0.008 (0.73)
–0.122 (0.002)
–0.119
Constr1970_1980
–0.123 (<0.001)
–0.005 (0.83)
–0.088 (0.03)
–0.083
Constr1992<
0.099 (<0.001)
0.099 (<0.001)
–0.085 (0.12)
0.104
CondMed
–0.109 (<0.001)
0.009 (0.63)
–0.072 (0.001)
–0.107
CondBad
–0.215 (<0.001)
–0.092 (0.04)
–0.197 (<0.001)
–0.225
ViewNo
–0.040 (<0.001)
–0.027 (0.03)
–0.017 (0.37)
–0.035
ViewBad
–0.098 (0.001)
<0.001 (0.99)
–0.002 (0.96)
–0.083
Cellar1
0.099 (0.001)
–0.002 (0.88)
0.092 (<0.001)
0.076
Cellar2
0.032 (0.001)
0.001 (0.98)
0.021 (0.84)
–0.005
Garden
0.049 (0.02)
–0.011 (0.73)
0.101 (0.16)
0.046
Terrace
0.043 (0.01)
0.049 (0.01)
0.037 (0.69)
0.041
NewApartment
0.046 (0.01)
–0.067 (0.001)
0.256 (<0.001)
0.047
District2
–0.109 (<0.001)
–0.035 (0.01)
–0.013 (0.64)
–0.053
District3
–0.130 (<0.001)
–0.104 (<0.001)
0.021 (0.54)
–0.088
District4
–0.229 (<0.001)
–0.188 (0.01)
0.093 (0.01)
–0.189
%MidIncome
0.671 (<0.001)
0.313 (0.05)
0.558 (0.01)
0.770
%HighIncome
0.136 (<0.001)
0.097
0.029 (0.61)
0.028
Adjusted R2 
0.838
0.839
0.705
0.875
N observations
4,362
873
872
4,362
Note: Significance level is in the parentheses.


===== PAGE 15 (score 95, numeric-paren 54, table-words 12, var-words 3) =====

115
Willingness to pay for accessibility under the conditions of residential segregation
APPENDIX 3 
OLS and Spatial Error models for sumbarkets
Submarket
Accessibility by car
Accessibility by public transport
TTBellecour
TTCityHall
Centrality 
Index
Accessibil-
ity Index
TTBellecour
TTCityHall
Centrality 
Index
Accessibil-
ity Index
OLS
Sainte-Foy-lès-Lyons,
84 obs.,
adj. R2 = 0.83 – 0.84, 
Max VIF = 3.25
–0.525
(0.02)
–0.381
(0.11)
0.022
(0.12)
0.006
(0.06)
–0.304
(0.05)
–0.203
(0.21)
0.013
(0.19)
0.003
(0.13)
Oullins,
100 obs.,
adj. R2 = 0.88 – 0.89, 
Max VIF = 3.35
0.098
(0.62)
0.306
(0.25)
–0.008
(0.39)
–0.003
(0.33)
0.307
(0.18)
0.396
(0.03)
–0.020
(0.08)
–0.005
(0.06)
Vénissieux,
97 obs.,
adj. R2 = 0.78 – 0.85, 
Max VIF = 5.56
–0.237
(0.17)
–0.478
(0.02)
0.022
(<0.001)
0.006
(<0.001)
–0.935
(<0.001)
–0.911
(<0.001)
0.061
(<0.001)
0.013
(<0.001)
Vaulx-en-Velin,
54 obs.,
adj. R2 = 0.66 – 0.68, 
Max VIF = 2.82
1.970
(0.19)
0.016
(0.99)
–0.014
(0.74)
–0.005
(0.69)
0.559
(0.54)
–0.355
(0.67)
–0.007
(0.87)
–0.004
(0.72)
Spatial Error
Sainte-Foy-lès-Lyons,
R2 = 0.88
–0.599
(0.01)
–0.427
(0.10)
0.027
(0.10)
0.007
(0.05)
–0.486
(0.01)
–0.268
(0.17)
0.021
(0.10)
0.005
(0.07)
Oullins,
R2 = 0.91
0.117
(0.59)
0.311
(0.27)
–0.008
(0.43)
–0.003
(0.35)
0.318
(0.18)
0.422
(0.02)
–0.022
(0.06)
–0.005
(0.05)
Vénissieux,
R2 = 0.87 – 0.89
0.284
(0.39)
–0.260
(0.45)
0.022
(0.04)
0.005
(0.11)
–0.947
(<0.001)
–1.162
(<0.001)
0.061
(<0.001)
0.014
(<0.001)
Vaulx-en-Velin,
R2 = 0.84
0.147
(0.94)
–0.562
(0.65)
0.019
(0.70)
0.003
(0.81)
0.685
(0.50)
0.499
(0.57)
–0.008
(0.86)
–0.005
(0.65)
Note: Significance level is in the parentheses.