http://usj.sagepub.

com/
Urban Studies
http://usj.sagepub.com/content/50/16/3205
The online version of this article can be found at:

DOI: 10.1177/0042098013484539
2013 50: 3205 originally published online 16 May 2013 Urban Stud
Marco Brambilla, Alessandra Michelangeli and Eugenio Peluso
Equity in the City: On Measuring Urban (Ine)quality of Life

Published by:
http://www.sagepublications.com
On behalf of:

Urban Studies Journal Foundation

can be found at: Urban Studies Additional services and information for

http://usj.sagepub.com/cgi/alerts Email Alerts:

http://usj.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.nav Reprints:

http://www.sagepub.com/journalsPermissions.nav Permissions:

What is This?

- May 16, 2013 OnlineFirst Version of Record

- Nov 15, 2013 Version of Record >>

at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
Equity in the City: On Measuring Urban
(Ine)quality of Life
Marco Brambilla, Alessandra Michelangeli and Eugenio Peluso
[Paper first received, August 2011; in final form, January 2013]
Abstract
In economic literature, the quality of life (QoL) in a city is usually assessed through
the standard revealed-preference approach, which defines a QoL index as the mone-
tary value of urban amenities. This paper proposes an innovative methodology to
measure urban QoL when equity concerns arise. The standard approach is extended
by introducing preferences for even accessibility to amenities throughout the city
into the QoL assessment. The QoL index is then reformulated to account for the
unequal availability of amenities across neighbourhoods. The more unbalanced the
distribution of amenities across neighbourhoods, the lower the assessment based on
the new index. This methodology is applied to derive a QoL index for the city of
Milan. The results show that the unequal distribution of amenities across neighbour-
hoods significantly affects the assessment of QoL for that city.
1. Introduction
The economic approach to measuring
urban quality of life (QoL) is based on the
work of Rosen (1979) and Roback (1982)
who, rather than assessing overall well-
being or happiness of households, measure
QoL indirectly, in terms of the monetary
value of local amenities. They define a QoL
index as the monetary value of a set of
amenities, where the amenities are location-
specific characteristics with positive or neg-
ative effects on individuals utility (Bartik
and Smith, 1987; Blomquist, 2006).
This paper focuses on a single city and
extends the previous methodology by pro-
viding a new measure to assess urban QoL
by explicitly taking into account equity
Marco Brambilla is in the DEFAP, Catholic University of Milan, Piazza Buonarroti 30, Milan,
20100, Italy. Email: mg.brambilla@polimi.it.
Alessandra Michelangeli (corresponding author) is in the Department of Economics, Management
and Statistics (DEMS), University of Milan-Bicocca, Piazza dellAteneo Nuovo 1, Building U6,
Milan, 20126, Italy. Email: alessandra.michelangeli@unimib.it.
Eugenio Peluso is in the Department of Economics, University of Verona, Vicolo Campofiore 4,
Verona, 37129, Italy. Email: eugenio.peluso@univr.it.
Urban Studies at 50
Article
50(16) 32053224, December 2013
0042-0980 Print/1360-063X Online
2013 Urban Studies Journal Limited
DOI: 10.1177/0042098013484539
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
concerns. Our methodology relies on the
hypothesis that a more equal distribution
of resources in the city is beneficial to over-
all QoL. As we will see later, there are sev-
eral arguments for equity in the city, some
of them used as a starting-point for devel-
oping a branch of planning and economic
literature. Regardless of the perspective
adopted for studying the effect of equity on
QoL, or more generally well-being, every-
one agrees that the opposite of equityi.e.
inequalityis detrimental to quality of life.
The concentration of poor or disadvan-
taged people in some areas of the city can
generate negative externalities and fuel
social discontent, eventually leading to
slums, delinquency, drugs, and other symp-
toms of social and economic marginalisa-
tion (see Glaeser et al., 2008). In our
framework, the priority for equity is cap-
tured by the mathematical properties of an
explicit evaluation function. The specific
form of the evaluation function can embed
either citizens preferences for equity
(where useful, investigated by suitable sur-
veys) or the ethical judgment of a city plan-
ner. The main assumption is that an
unequal availability of amenities within the
city has a negative impact on the evaluation
function; in the same way as income
inequality generates a loss in social welfare
according to Atkinsons (1970) approach to
inequality measurement. Under this
assumption, a new QoL index is obtained
by discounting the Roback (1982) index
through a multiplicative correction term.
The more unbalanced the distribution of
amenities across neighbourhoods, the lower
the correction term. More precisely, the
correction term is obtained as the sum of
unidimensional inequality indices, account-
ing for the dispersion of each amenity
within the city, plus a residual term sum-
marising any correlation among the distri-
bution of amenities. This formulation can
be used to disentangle the contribution of
the dispersion of each amenity to the over-
all index from the joint effect of the ame-
nities. The correction term depends on as
many parameters as the number of ame-
nities under examination. Each parameter
registers the aversion to the unequal avail-
ability of the corresponding amenity within
the city. The model is therefore sufficiently
flexible to allow for a specific degree of
aversion to the unequal availability of each
amenity.
One point requires further explanation.
Our analysis is based on two different com-
ponents: the Roback (1982) spatial model
and an evaluation function expressing the
preferences for equity. In the former, the
representative agent looks for the most con-
venient location given the distribution of
amenities across neighbourhoods. The dif-
ferent amenities available at the equilibrium
are capitalised in housing prices and the
representative agents utility is equalised
within the city. This outcome is logically
independent of the fact that it might be
possible to increase or re-allocate some
amenities within the city, improving the
well-being of the representative citizen. This
effect is captured by the correction term
depending on the evaluation function. For
example, in the short run, agent location
decisions depend on the current availability
of green areas within the city, which also
affects his well-being at the equilibrium.
However, if new green areas are developed,
especially in neighbourhoods where they
are lacking, overall well-being in the city
increases. These points of view are recon-
ciled by first computing a Roback QoL
index based on the citizens evaluation of
amenities, then adjusted via the correction
term sensitive to their unequal availability.
The suggested methodology is illustrated
using data for the city of Milan over the
period 200408 regarding the availability of
education, green areas, recreational activi-
ties, commercial facilities, public transport
3206 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
and socio-demographic characteristics. By
taking into account the uneven availability
of amenities within the city, the Roback
index is reduced by 28 per cent.
The methodology is shown in section 2.
Section 3 presents the empirical application
to Milan with a discussion of the data and
variables and a descriptive analysis of the
city neighbourhoods. The econometric spe-
cification of the hedonic function is also
discussed and the results are illustrated in
terms of amenity prices and QoL assess-
ment. Section 4 concludes and suggests
some potential extensions of our approach,
paving the way for future research.
2. Theoretical Framework
This section first reviews the Roback (1982)
index, then shows how to measure the avail-
ability of amenities within the city and,
thirdly, shows how to obtain a new QoL
index accounting for the unequal availabil-
ity of amenities across neighbourhoods.
2.1 The Roback (1982) QoL Index
Let a 2 <
k
+
be the vector of the average
quantities of k amenities in a given city.
The index developed by Rosen (1979) and
Roback (1982) consists of the weighted
sum of the values of the k amenities of the
city, where the weights are the implicit
prices associated with the amenities
QoL =

k
j =1
p
j
a
j
1
The implicit price p
j
for j = 1, ., k is esti-
mated through housing and wage hedonic
regressions. Actually the coefficients in the
estimated hedonic regression do not only
reflect consumer preferences, but also factors
that determine production. However, in the
case of a fixed supplied amenity, the hedonic
price reflects the preferences of the marginal
person just indifferent between buying a
house with and without amenity. For ame-
nities that vary more continuously through-
out a city, a hedonic regression is likely to
capture something much closer to mean pre-
ferences (Bayer and McMillan, 2008, p. 229).
The implicit price is the sum of the housing
price differential and the negative of the
wage differential. In other words, the eco-
nomic value of a local amenity is determined
by the housing price individuals are willing
to pay and the wage they are willing to accept
to locate in a given city. The idea underlying
this approach is that people will accept lower
wages and/or greater housing prices in an
area with desirable amenities, but require
greater wages and/or lower housing prices in
an area with less attractive amenities. As
Roback acknowledges
associating the implicit marginal price with
mean quantity of the amenity provides an
approximated value of the amenities because
the prices of infra-marginal units are differ-
ent from the marginal prices. Such an expen-
diture computation in (1) merely shows the
order of magnitude of the expenditure in the
average budget (Roback, 1982, p. 1274).
In what follows, the focus is on a single
city,
1
computing a QoL index for the whole
city that accounts for the uneven availabil-
ity of amenities across its neighbourhoods.
Focusing on a single city has three main
consequences.
First, the implicit prices of amenities in
(1) are derived solely from the hedonic
housing price equation. Wages are ignored
since they are assumed to be determined for
the citys labour market as a whole without
variation within the city. Actually, several
studies carried out on American cities show
that wages may vary slightly within a city.
According to Bartik and Eberts (2006), for
example, the wages of identical workers
EQUITY IN THE CITY 3207
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
decline about 1 per cent for each additional
mile the job is located from the central busi-
ness district (CBD). Even supposing that
wages can vary within a city, it is not easy to
measure such a phenomenon, since the
neighbourhood where individuals live is not
necessarily where they work. In addition,
the lack of data on employment precludes a
consideration of city-wide wage variations.
Secondly, only the prices of amenities
that vary within the city can be identified.
We ignore several variables usually consid-
ered in intercity analysis (for example, cli-
mate, altitude).
A third potential problem with intracity
analysis relates to spatial sorting on unob-
servables (Gyourko et al., 1999). This occurs
when high-quality housing units are located
in the best city neighbourhoods and the fac-
tors determining the high quality of houses
are unobservable. The value of QoL will be
overestimated in the nice neighbourhoods.
This point will be returned to in section 4.
The conventional approach is now
extended to quantify the availability of
amenities for city inhabitants living in dif-
ferent neighbourhoods.
2.2 Amenities and Their Availability
Let us consider a city exogenously parti-
tioned into n neighbourhoods. Each neigh-
bourhood i =1, . . . , n is described by a
vector a
i
containing the values of the k ame-
nities. The element a
ij
2 <
+
indicates the
level of the amenity j in the neighbourhood
i. The information on the distribution of
amenities in the city is then summarised by
a n x k matrix, where each line corresponds
to a neighbourhood and each column to an
amenity
A=
a
11
. . .a
1k
a
n1
. . .a
nk
_
_
_
_
Suppose that an individual lives in a neigh-
bourhood with few amenities or none at all.
He can benefit from amenities located in the
surrounding neighbourhoods. Therefore,
the overall quantity of the amenities avail-
able is the sum of the amenities where the
individual dwells, plus a term indicating
their presence in the surroundings, whose
accessibility is a function of the distance
between the neighbourhood we are consid-
ering and its adjacent neighbourhoods (see
Figure 1).
In formal terms, starting from A, a fur-
ther n x k matrix Z is obtained, whose gen-
eric term z
ij
indicates the overall availability
of amenity j for individuals of neighbour-
hood i. The element z
ij
is obtained by adding
to a
ij
the availability of the amenity j in the
neighbourhoods bordering i. Defining S(i)
the set of neighbourhoods adjacent to i
(with i 2 S(i)), we obtain
z
ij
=

s2S(i)
a
is
f (d
is
) 2
where, a
sj
is the value of amenity j in the
neighbourhood s bordering i; f d
is
is the
value of a continuous and non increasing
function f : <
+
! 0, 1 defined on the dis-
tance d
is
between the neighbourhoods i and
s.
2
Let z
j
denote the average level of amenity
j in the city and z = z
1
, ::, z
k
the vector con-
taining the means of the k amenities in the
city. We reformulate the Roback (1982)
index as follows
il
d
Figure 1. Distance between district centroids.
3208 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
QoL =

k
j =1
p
j
z
j
3
where, the vector of the average quantities a
is replaced by z.
We call this index the QoL index adjusted
for the available amenities. A procedure to
correct this index accounting for inequality
is now presented.
2.3 The Equity-adjusted QoL Index
The idea of just city is enshrined in both
planning and economic literature. The
equity approach developed in planning lit-
erature requires that urban amenities and
public services must be evenly available in a
way such that
everyone receives the same public benefit,
regardless of socioeconomic status, willing-
ness or ability to pay, or other criteria; resi-
dents receive either equal input or equal
benefit (Talen, 1998, p. 24).
An alternative approach (needs-based), fol-
lows a compensatory criterion that subor-
dinates the distribution of facilities and
services to the different needs of the popu-
lation living in different neighbourhoods of
the city.
3
In urban economics literature, aiming at
evenly available amenities within the city is
recommended by Berliant et al. (2006),
who determine the optimum number and
location of public facilities through a gen-
eral equilibrium analysis. They prove the
existence of an equilibrium characterised
by a dispersed and homogeneous distribu-
tion of public facilities across locations.
They also argue that an equal treatment
identical-provision optimum may be justi-
fied by the equal protection clause of the
US Constitution, federal law or by state
constitutions. On the efficiency side,
Benabou (1993) shows that stratification
can create ghettos and can even bring
about the complete collapse of the citys
productive capacity. Finally, from a social
welfare perspective, the equal availability of
local facilities mitigates well-being inequal-
ity (Aaberge et al., 2010) and promotes
equality of opportunities in the sense of
Roemer (1998) and van de Gaer (1993).
Moving from these different equity con-
cerns to an analysis of QoL, a measure of
QoL is now derived. This measure is able
to disentangle the different relevance, if
any, given to the unequal availability of
each amenity. For example, the unequal
availability of shopping facilities can have a
higher impact in lowering QoL than the
identically unequal availability of cultural
sites. Let W(Z) be the function evaluating
the distribution of the k amenities between
the n city neighbourhoods.
4
The function
W is assumed as additive
W(Z) =
1
n

n
i =1
w(z
i
) 4
where, w(z
i
) is the value taken by the
increasing and concave function
w : <
k
+
! < summarising the assessment
of the QoL in the neighbourhood i,
endowed with the vector of amenities z
i
.
It is well known (see Weymark, 2006)
that, under inequality aversion, the value
W(Z) is less than or equal to the value guar-
anteed by an even availability of the same
amount of amenities across neighbour-
hoods. If
~
W =W(Z)is the evaluation of the
current distribution, due to continuity, it is
possible to define a scalar q 2 0, 1 depend-
ing on Z and W, such that
~
W =W(Z) =
1
n

n
i =1
w(z
i
) =W(q z) 5
In the same spirit as Atkinson (1970), the
elements of the vector qz are called equally
distributed equivalent amenities (EDEAs).
EQUITY IN THE CITY 3209
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
This equation (5) means that, if the propor-
tion q of the average amount of amenities is
uniformly available across neighbourhoods,
the evaluation function W(q z) reaches the
same value W(Z) associated with the actual
distribution.
Accounting for inequality, the index
adjusted for the available amenities (3) is
then modified as follows
Definition 1. For any Z, the equity-adjusted
urban quality of life index QoL
q
(Z; W) is
QoL
q
(Z; W) =q

k
j =1
p
j
z
j
6
where, p
j
is the implicit price of the amenity j
and the equally-distributed equivalent ame-
nities vector qz is implicitly defined by (5).
Instead of assessing the value of z repre-
senting the average quantity of amenities
available in the city as in (3), the equity-
adjusted QoL index (6) corrects it through
a multiplicative factor q. The vector qz
consists in the multidimensional extension
(Tsui, 1995; Weymark, 2006) of the equally
distributed equivalent income used in
inequality literature (Atkinson, 1970) and
of the certainty equivalent used in risk
theory (Pratt, 1964). The more unbalanced
the distribution of amenities, the lower the
correction term q. In the Appendix, we
show how to derive q from the evaluation
function (5). The proposed methodology
can be used to disentangle the contribution
of each amenity to the value of q and then
to the overall QoL index. The correction
term q is derived through a parametric
inequality index embedding preferences for
an equal availability of the k amenities
within the city. The k parameters measure
the level of inequality aversion of either citi-
zens or a city planner on each dimension.
Identifying the origin of the aversion to
inequality is beyond the aim of this paper.
Our aim is to show how preferences for
equity can be introduced in QoL assess-
ment, without recommending a specific
equity perspective. In this respect, our
attempt offers a pioneering contribution to
both urban and inequality literature.
5
To
proceed with the empirical investigation, a
pragmatic procedure is adopted to calibrate
inequality aversion, exploiting the available
information on the willingness to pay for
each amenity, estimated through a hedonic
model referred to the housing market. The
k parameters indicating aversion to the
unequal availability of the different ame-
nities will be set equal to their respective
weight on the citys monetary assessment of
quality of life.
3. Empirical Application
In this section, we employ the previous
model to assess QoL in Milan, the largest
city in Italy after Rome.
6
In addition to
being the biggest industrial city in Italy, it is
also a historical city with a significant assort-
ment of churches, buildings and monu-
ments mainly inside the Mura Spagnole, the
city walls that border the historical city
centre. Despite all these positive aspects,
other factors do not positively impact on
quality of life. Some neighbourhoods have
experienced a gradual process of urban
decay and more and more Italian residents
have abandoned these areas. Housing prices
have decreased and these neighbourhoods
have frequently attracted newcomers to
Milan. The next section describes the infor-
mation we have collected to capture these
phenomena and, more generally, the main
variables affecting the quality of life in the
different neighbourhoods of Milan.
3.1 Data and Variables
As we showed in relation to the Roback
(1982) QoL index in (1), overall QoL
3210 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
depends on the set of amenities considered
when implementing the analysis. For the
purpose of this study, several data sources
are combined into a single dataset that con-
tains detailed information on housing and
city characteristics. Data on residential
housing transactions were taken from the
Osservatorio del Mercato Immobiliare
(OMI) managed by a public agency (the
Agenzia del Territorio). Transactions refer
to 55 neighbourhoods identified by the
OMI for a period of 5 years, from January
2004 to December 2008.
7
Each neighbour-
hood is internally homogeneous in terms of
socioeconomic and urban characteristics,
making it likely that prices of houses located
within a neighbourhood move together. In
addition to the housing market value, the
dataset provides a detailed description of
structural attributes, such as total floor
space, age of the building in the year of sale,
number of bathrooms, whether the housing
unit needs to be renovated, whether the
housing unit has independent heating, the
floor above street level, presence of an eleva-
tor or a garage, and build quality.
Neighborhood-level data on amenities
and socioeconomic conditions were taken
from public authority records. They include
information on six important aspects of
quality of life: environmental characteris-
tics, public transport, education, shopping
facilities, recreational activities, and socioe-
conomic characteristics. Table A2 in the
Appendix sets out a full list of variables
used in our analysis with their sources. Of
course, other variables that cannot be so
readily observed may contribute to the
quality of life as presented in this paper. In
addition, depending on the revealed prefer-
ences of residents over a limited bundle of
amenities, our QoL index could not prop-
erly account for the preferences of those res-
idents who give priority to other (omitted)
amenities. However, this paper aims to
show the potential of our methodology,
carrying out an empirical analysis which is
as rigorous as possible.
The environmental dimension is proxied
by the green areas relative to the area of the
neighbourhood (Green); public transport is
represented by the number of metro stations
(Transport); shopping facilities (Shopping
facilities) are proxied by the number of
supermarkets, discount stores and malls per
10 000 inhabitants; the recreational dimen-
sion (Cultural) is proxied by the number of
cinemas, theatres, museums, art galleries,
academies of music and libraries per 10 000
inhabitants.
The socioeconomic dimension (Ethnic)
is based on the ratio of Italian/foreign resi-
dents. More precisely, the Ethnic variable
for the neighbourhood i is constructed as
Ethnic
i
=
It
i

s2S(i)
Imm
s
~
d
2
is
where, It
i
is the number of Italians living in
the neighbourhood i. In the denominator,
we add to the immigrants living in the
neighbourhood, Imm
i
, those living in the
surrounding neighbourhoods. The weight
associated with immigrants of each adjacent
neighbourhood s 2 S(i) is set to be equal to
1
~
d
2
is
, with
~
d
2
is
=max(1, d
is
) and roughly approx-
imates the probability that Italian residents
in neighbourhood i interact with foreigners
in neighbourhood s. In this case, our
assumption of preference for equity corre-
sponds to aversion to residential segregation.
The resulting unidimensional index for the
variable Ethnic can be safely interpreted in
terms of a segregation index of the exposure
class (see Reardon and OSullivan, 2004). In
this spirit, we consider as foreign residents
(Imm) the ethnic groups most often subject
to discrimination, such as African, South
American and some Asian communities.
Other immigrantsfrom, for example,
North America or western Europehave
EQUITY IN THE CITY 3211
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
become invisible in Italy
8
and are included
with Italian residents.
Education is proxied by the proximity to
the nearest university (Proximity_University).
Unfortunately, we do not have information
on other variables for the quality of educa-
tion, such as the percentage of pupils moving
up to a higher class or parameters for class-
room and/or building facilities.
9
We do have
information on the degree of availability of
different educational levels, such as the
number of primary and secondary schools,
both public and private, in the neighbour-
hood. We also have data on early years of
educationi.e. the number of nursery
schools and pre-schools. We tried alternative
specifications including these variables but
none of these turned out to be statistically
significant. There are at least two reasons for
this result. First, the number of schools avail-
able in the neighbourhood is a rough proxy
of education services and is unable to capture
the quality of education services. Secondly,
the variability of some of these covariates is
quite modest across neighbourhoods.
In addition to these amenities, the
Euclidean proximity of each neighbour-
hood to the city centre is included, in order
to handle the problem of the spatial sorting
on unobservables described in section 2.1.
Furthermore, the Euclidean proximity to
the city centre enables verification of the
hypothesis of a monocentric structure of
the urban area modelled by Alonso (1965)
and Muth (1969). The urban pattern of
Milan is clearly identifiable by the old inner
ring of inland waterways (Navigli) designed
by Leonardo da Vinci. Outside the former,
a second ring comprises the Mura Spagnole.
The most important historical monuments
(Duomo, Castello Sforzesco, Royal Palace,
etc.) and historical buildings of residential
use are located within the two rings, beyond
which, as far as the confines of the munici-
pal territory, large neighbourhoods spread
out.
10
Descriptive statistics are set out in Table 1.
Amenity statistics enshrine the availability of
amenities in the bordering zones, calculated
by using the distance function f d
is
=
1
~
d
2
is
.
Following Marans (2003), the distance is
expressed in miles: amenities located in sur-
rounding neighbourhoods at a distance less
than or equal to a mile are added to that of
the initial neighbourhood entirely, while
those further than a mile are divided by d
2
is
.
The specification of f can be motivated in
this case by the current practice adopted in
gravitation models (White, 1983; Batten and
Boyce, 1986; Wong, 1993).
The average value of the 2592 properties
sold over the 200408 period was e403 288.
The average property had 95.72 square
metres of total floor space and was 48 years
old at the time of sale. Each neighbourhood
had on average 12 per cent of greenery in the
urban area, but with substantial differences
from 1 per cent in the Ronchetto Chiaravalle
Ripamonti neighbourhood in the south-west
to 25 per cent in the off-centre neighbour-
hood of Monza Precotto Gorla in the north.
The number of metro stations also varies
greatly: from 0 to 13 stations. Each neigh-
bourhood had on average 9.44 shopping
areas and five cultural sites per 10 000 inha-
bitants. Finally, the average ratio of foreign to
Italian residents in the neighbourhood was 9
per cent, the minimum percentage being 3.22
and the maximum 21.26.
3.2 Estimated Implicit Prices
To obtain the full implicit prices of location-
specific amenities given in (6), we estimate a
reduced form of the housing price hedonic
equation
ln p
hit
=b
0
+b
1
X
hit
+b
2
A
it
+h
hit
7
where, ln p
hit
is the price of housing unit h in
neighbourhood i at time t; X
hit
is a vector of
housing characteristics; A
it
is a vector of
3212 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
available amenities in neighbourhood i; h
hit
~
N(0, s
2
n
):The structural characteristics in X
hit
are listed in Table A2. In addition to these
characteristics, the proximity of each neigh-
bourhood to the city centre and year dummies
are included in X
hit
to account for unobserva-
ble factors and time fixed effects respectively.
3.3 Results
Table 2 presents the results obtained from
estimating (7) by OLS. Robust standard
errors are used with clustering at neighbour-
hood level in order to allow for within-
neighborhood correlation. All in all, the
housing variables used in the model account
for about 87 per cent of the variance of the
logarithm of price. The stability of the
regression model was verified using pairwise
Chow tests on adjacent sub-periods. The
hypothesis of no structural change is not
rejected with the exception of the sub-
period 200607. Therefore our model
appears to be sufficiently stable. The amenity
Table 1. Summary statistics for the variables
Variable Mean S.D. Minimum Maximum Unit
Housing value 403 288 409 900 80 000 7 000 000 Euro
Market_p 0.304 0.460 0 1 Dummy
Offer_p 0.446 0.496 0 1 Dummy
Estimated_p 0.251 0.434 0 1 Dummy
Amenities
Transports 4.36 3.31 0 12.90 Continuous
Proximity_university 4.61 1.62 0 6.60 Km
Green 12.04 5.24 1 25 Percentage
Cultural 5.00 10.74 0.06 70.43 Continuous
Shopping_facilities 9.44 4.96 1.34 21.64 Continuous
Ethnic 9.14 3.75 3.22 21.26 Percentage
Housing characteristics
Total floor area 95.72 46.01 19.00 452 Square metres
Number of bathrooms 1.36 0.63 1 6 Discrete
To be renovated 0.042 0.200 0 1 Dummy
Heating 0.124 0.330 0 1 Dummy
2nd floor (or higher) 0.817 0.386 0 1 Dummy
Low_cost 0.538 0.498 0 1 Dummy
Medium_cost 0.428 0.494 0 1 Dummy
Luxury 0.032 0.178 0 1 Dummy
Parking 0.019 0.067 0 1 Dummy
Elevator 0.819 0.384 0 1 Dummy
Age 48.20 21.90 1 205 Discrete
Control variables
City centre 0.15 0.35 0 1 Dummy
Proximity_city centre 4.50 2.20 0.1 9.3 Km
Year 2004 (ref.) 0.14 0.35 0 1 Dummy
Year 2005 0.23 0.42 0 1 Dummy
Year 2006 0.19 0.39 0 1 Dummy
Year 2007 0.21 0.41 0 1 Dummy
Year 2008 0.20 0.40 0 1 Dummy
EQUITY IN THE CITY 3213
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
coefficients are statistically significant; to
quantify their relative importance in our
specification, standardised beta coefficients
are set out in column 3.
11
According to this
criterion, the most important amenity is
Green since an increase by one standard
deviation in this variable implies an increase
of 0.161 standard deviation in the value of
the housing unit. One possible explanation
of the statistical relevance of this variable is
that it measures not only the size of the avail-
able green areas, but also the facilities which
are often located within gardens and parks
(playgrounds for children, bicycle lanes and
sports centres). The next amenity in terms of
importance is Cultural (0.156). In this case,
Table 2. Hedonic regression
Variable Coefficient t Standard b coefficient
Intercept 10.600 273.90
Offer_p 0.062 4.51 0.048
Estimated_p 0.007 0.51 0.005
Amenities
Transport 0.007 4.21 0.041
Proximity_university 0.004 1.05 0.011
Green 0.001 2.02 0.161
Cultural 0.009 11.72 0.155
Shopping_facilities 0.003 2.31 0.025
Ethnic 0.008 5.84 0.048
Housing chacteristics
ln(size) 0.940 67.07 0.662
Second bathroom 0.048 3.65 0.035
Third bathroom 0.093 3.49 0.027
To be renewed 0.110 4.85 0.035
Heating 0.012 0.86 0.006
2nf floor (or higher) 0.017 1.53 0.010
Medium cost 0.029 2.73 0.023
Luxury 0.116 4.21 0.032
Parking 0.154 3.33 0.023
Elevator 0.068 5.41 0.041
ln(age) 0.006 1.51 0.018
proximity 0.0003 3.48 0.123
proximity squared 0.0001 1.83 0.077
Control variables
City centre 0.104 4.16 0.059
Year 2005 0.040 2.72 0.027
Year 2006 0.108 7.08 0.068
Year 2007 0.101 6.09 0.663
Year 2008 0.104 6.03 0.067
Number of observations 2592
F(26; 2565) 672.8400
Prob . F 0.0000
R
2
0.8721
Adjusted R
2
0.8708
3214 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
the importance could be explained by the
location of specific theatres and museums in
ancient buildings where individuals appreci-
ate aesthetic and artistic value. For example,
the Museum of Ancient Art is in the Castello
Sforzesco, probably the most famous build-
ing in Milan, together with the Duomo and
La Scala Opera House. The other standar-
dised beta coefficients are 0.048 for Ethnic,
0.042 for Transport, 0.025 for Shopping facil-
ities and 0.011 for the proximity to the near-
est university (Proximity_university).
Full implicit prices for each amenity are
shown in Table 3, with Cultural, Ethnic and
Transport showing the highest prices in
absolute value. The hedonic price for an
additional cultural site per 10 000 inhabi-
tants in the areal unit is e2839; increasing
the Italian/foreign ratio by one leads to an
increase of e2500 in the value of the aver-
age housing unit and an additional metro
station provides a benefit of e2470. We
included distance from the nearest univer-
sity and it turns out that reducing the dis-
tance by 1 km increases housing unit value
by e1377. The estimated implicit price for
an additional shopping facility per 10 000
inhabitants is e1051 and the hedonic price
for public green areas is e 612 for one more
percentage point.
Column 2 in Table 3 sets out the values of
e, given by (A6), which corresponds to the
contribution of each amenity to the determi-
nation of the overall QoL index adjusted by
the availability of the amenities defined by
equation (3). These values are represented in
the pie chart in Figure 2 where it is evident
that variables related to social dimension
(Ethnic) and leisure (Cultural) play the most
important role.
According to the model presented in sec-
tion 3 and detailed in the Appendix, the
weights s= s
1
, ::, s
k
assigned to the set
of amenities are calculated using the esti-
mated implicit prices and the average quan-
tities of amenities in the city according to
equations (A5) and (A6). The third column
in Table 3 shows that these weights are
fairly similar and that the lowest is assigned
to the socioeconomic amenity Ethnic. This
implies that aversion to the unequal distri-
bution of immigrants across neighbour-
hoods will be higher than to the unequal
distribution of the other amenities.
The last column in Table 3 shows the
values of the equally distributed equivalent
share g
j
associated with each amenity j. As
stated earlier, it corresponds to the share of
the city-wide average value of the amenity j
so that, if equally distributed throughout all
Table 3. Hedonic prices and QoL indexes
Variable p
j
(e) e
j
=
p
j
z
j

k
l =1
p
l
z
l
[A6] s
j
=
1e
j
k1
[A5] l
j
=
1
z
j
1
n

n
i =1
z
s
j
ij
_ _1
s
j
[A2]
Transport (e/station) 2 475 0.151 0.1698 0.5418
Prox_univ (e/km) 1 377 0.089 0.1822 0.8608
Green (e/ +1%) 612 0.103 0.1794 0.5683
Cultural (e/unit) 2 839 0.198 0.1603 0.3339
Shop_fac (e/unit) 1 051 0.139 0.1722 0.8671
Ethnic (e/unit) 2 500 0.320 0.1360 0.9300
QoL index (equation (3)) 71 473
Correlation term k 1.0909
Correction term u 0.72
Equity-adjusted index QoL
u
51 460
EQUITY IN THE CITY 3215
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
neighbourhoods, it would provide the same
value of the evaluation function (4) assessed
with the actual distribution of j, as claimed
by equation (5). Its value depends both on
the degree of aversion to the unequal distri-
bution of amenity j and its actual distribu-
tion across neighbourhoods. The highest
value of g
j
is for Ethnic (0.93), which shows
a limited degree of residential segregation
in Milan, followed by Shopping facilities
(0.86) and Proximity to university (0.86).
The lower g
j
found for Cultural (0.33)
Transport (0.54) and Green (0.56) roughly
indicate an unequal availability of these
amenities across city neighbourhoods.
Finally, the interaction term k is calculated
to be slightly bigger than 1 (1.09),
12
imply-
ing that the joint effect of amenity distribu-
tions positively contributes to the overall
equity-adjusted QoL index.
The last step backward defines the extent
of the reduction of the QoL index adjusted
for the amenity availability shown in (3).
This value amounts to e71 473. The calcu-
lations carried out show that the overall
correction term q amounts to 0.72 and
therefore the QoL index accounting for
equity is equal to e51 460. This result does
not differ so much from the benchmark
case where e
j
is 1/6 for all j and inequality
aversion is the same for each amenity. The
correction term q reduces from 0.28 to
0.26 and the adjusted QoL value rises to
e52 246.
3.4 Discussion
This result implies that the uneven availabil-
ity of amenities within the city reduces the
index adjusted for the availability of ame-
nities (3) by 28 per cent. Increasing the level
of amenities improves quality of life but
their even availability also matters. Figure 3
shows the level of QoL by neighbourhood
computed according to (3).
13
This map
confirms the monocentric shape of Milan
where QoL levels are much higher in the
city centre and decrease as the distance from
the centre increases. All neighbourhoods in
the city centre have a QoL value near to or
far exceeding e100 000, with the exception
of neighbourhood 6Castello, Melzi dEril,
Sarpi with a value of e74 775. This lower
value is due to the numerous settlement in
the past 10 years of wholesalers trading in
clothing and leather goods, essentially com-
prising a Chinese community. Moreover, in
the district there is friction between resi-
dents and wholesalers because of the
Figure 2. Contribution of amenities to the overall QoL index adjusted for the availability of
amenities.
3216 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
continuous loading and unloading of
goods, causing traffic problems and endan-
gering pedestrians. The lowest value
(e30 096) is for neighbourhood 55
Quarto Oggiaro, Roserio, Amorettiwhich
sprang up in the 1950s to house workers
from southern Italy. In the course of time,
this neighbourhood developed a reputation
for organised crime, dilapidated housing
and a high percentage of illegal immigrants
(of the 4000 apartments made available by
public housing, 700 are occupied illegally).
The relationship between the value of
QoL and income across neighbourhoods is
examined. The Pearson correlation coeffi-
cient is 0.846 implying that QoL is strongly
positively related to income. This result is in
line with the study of Brueckner et al. (1999)
regarding the importance of amenities in
driving the location choice of rich individu-
als towards the best-endowed neighbour-
hoods, rises rapidly with income.
We conclude this section with a concern
about whether these empirical results would
still be meaningful if the needs-based criter-
ion of equity in the city, cited in section 2.3,
was adopted. Suppose for instance that
households comprising people of over 65
years of age are concentrated in a few city
neighbourhoods. In this case, it would be
reasonable to concentrate amenities such as
healthcare services in these neighbour-
hoods. To investigate this point, we looked
at the distribution of two kinds of house-
holds with specific traits able to shape their
preferences (households with children and
people over 65) in Milan. It turns out that
these households are evenly distributed
across city neighbourhoods (their Gini con-
centration index is respectively 0.056 and
0.068) so we can conclude that our results
hold up against the preference heterogeneity
induced by the demographic composition
of households.
0 - 40 000
40 000 - 75 000
> 75 000
52
48
49
46
47
50
51
53
23
25
55
26
24
27
30
28
29
32
31
33
35
34
21
20
36
37 38
19
39
40
42
41
54
43
44
45
18
17
16
10
9
15
13
5
14
12
11
22
6 7
4
3
2
1
8
1 km
Figure 3. QoL across Milan neighbourhoods. Key: see Table A1 for key to names of
neighbourhoods.
EQUITY IN THE CITY 3217
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
4. Concluding Remarks
This paper is an original attempt to bridge
the gap between the urban economics and
inequality measurement literature. Starting
from the premise that these two fields of
economics share a multidimensional view
of QoL, we propose an innovative metho-
dology to assess urban QoL when equity
concerns arise. Without recommending any
specific equity perspective, we show how
preferences for equity can be introduced
into QoL assessment through a simple list
of parameters which generate a correction
term for the Roback QoL index adjusted
for amenity availability. Our empirical
investigation for Milan indicates the quite
significant impact of inequality on the QoL
index for the whole city. To disentangle
this result, we assessed QoL across neigh-
bourhoods, showing significant differences.
Milan has a monocentric shape like other
major European cities both in terms of
income and QoL values. These two vari-
ables are strongly positively related across
Milans neighbourhoods. Better-endowed
districts attract wealthier households. It fol-
lows that to decrease stratification in the
city, improving efficiency and equalising
opportunities and life-chances in the spirit
of Massey and Denton (1996), policies
favouring a more even availability of ame-
nities should be promoted. In this perspec-
tive, our methodology has a wide range of
applications, from simulating the effects of
changes in the provision of public goods on
QoL, to the analysis of poverty and stratifi-
cation at the urban level. It can also be
used to assess the effects of gentrification
(Helms, 2003; Lees, 2008) or urban renova-
tion policies (Barthelemy et al., 2007).
Finally, our approach could be extended on
the basis that certain groups might be
better clustered for equity reasons (i.e. the
elderly or families with children). In this
case, the correction term applied to the
QoL indicator should account for the
degree of association between the distribu-
tion of these categories of people and the
spatial provision of specific amenities. This
constitutes a promising avenue for future
research.
Notes
1. A large number of studies have used data
for a single city focusing in particular on
environmental issues (air or water pollu-
tion, proximity to noxious sites, etc.). See,
for example, Kohlhase (1991); Michaels and
Smith (1990).
2. In our application, for the sake of simpli-
city, we use the Euclidean distance between
the centroids of the neighbourhoods. Of
course a more refined technique would be
to use commuting time and mobility costs,
where these data are available.
3. At the end of the empirical section, we
show that the results for Milan also prove
to be robust in the light of this second per-
spective. See Lucy (1981) for a taxonomy of
equity criteria in urban planning and Rusk
(2003) for a long-term analysis of US cities.
4. To simplify the presentation, in what fol-
lows we simply call amenities the ame-
nities adjusted for their availability.
5. Indeed, in the body of literature on
inequality and welfare measurement, the
traditional approach based on the prefer-
ences of a social planner has only recently
been questioned by alternative approaches
based on individual preferences (see
Fleurbaey and Maniquet, 2011).
6. The total population was 1 295 339 inhabi-
tants in 2008, within an area of about 183
square km.
7. See Table A1 in the Appendix for the list of
districts with their population size.
8. As in other Western countries; see for
example Pan Ke Shon (2010) for France.
9. By classroom facilities we mean, for exam-
ple, modern teaching aids, air conditioned
rooms, spacious rooms and neat and clean
classrooms. The building facilities are, for
example, recreation and gym facilities,
3218 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
high-speed Internet access, an extensive
library and computer lab facilities.
10. Michelangeli and Zanardi (2009) provide
empirical evidence for a monocentric shape
of the residential housing market in Milan.
Gonzales (2009, sec. 3) describes the histor-
ical and political roots of the monocentric
structure of Milan. More recently Veneri
(2013) accounts for the transformation into
an increasingly polycentric space of the
Milan metropolitan area. Extending our
analysis to such a larger area would require
an alternative empirical specification of the
hedonic equation (7).
11. The standardised beta coefficient quantifies
how many standard deviations change the
value of a house when each control variable
is increased by one standard deviation.
12. If k were equal to 1 there would be no joint
effect of amenity distributions on QoL index.
13. To compute the QoL index adjusted for the
available amenities for neighbourhood i,
with i = 1, ., n, equation (3) is modified
as follows
QoL
i
=

k
j =1
p
j
z
ij
14. We do not explore these properties and
invite the reader interested in axiomatics to
consult Tsui (1995).
15. The reader will note that 1 l
j
is exactly
the famous Atkinson (1970) inequality
index: so l
j
approaches 1 when the amenity
j is evenly available within the city; while l
j
approaches 0 when inhabitants of only few
neighbourhoods are neatly advantaged.
16. See also Brambilla and Peluso (2010).
17. This issue is usually solved in multidimen-
sional inequality literature by resorting to
sensitivity analysisi.e. by showing how the
results stand up to different numeric values
for the parameters.
18. See www.lisdatacentre.org/data-access/key-
figures/download-key-figures/.
Acknowledgements
We gratefully acknowledge the Osservatorio del
Mercato Immobiliare for data on housing trans-
actions; the Statistics Bureau of the Municipality
of Milan joint with the Department of Statistics
of the University of Milano-Bicocca for data on
income and demographic variables; Luca Stanca
for data on public transportation. We would like
to thank the editors for their time and valuable
remarks. We thank also Rolf Aaberge, Francesco
Andreoli, Michel Le Breton and Vito Peragine for
useful discussions. The usual disclaimer applies.
Funding
Financial support by the Italian Ministry of
University and Research and the Institute for
Economic Research on Firms in Vicenza is grate-
fully acknowledged.
References
Aaberge, A., Bhullera, M., Langrgena, A. and
Mogstad, M. (2010) The distributional
impact of public services when needs differ,
Journal of Public Economics, 94, pp. 549562.
Abul Naga, R. H. and Geoffard, P.-Y. (2006)
Decomposition of bivariate inequality by
attributes, Economic Letters, 90, pp. 362367.
Alonso, W. (1965) Location and Land Use:
Toward a General Theory of Land Rent. Cam-
bridge, MA: Harvard University Press.
Atkinson, A. (1970) On the measurement of
inequality, Journal of Economic Theory, 3, pp.
244263.
Barthelemy, F., Michelangeli, A. and Trannoy, A.
(2007) La renovation de la Goutte dOr est-
elle un succe`s? Un diagnostic a` laide dindices
de prix immobilier, Economie et prevision,
180/181, pp. 107126.
Bartik, T. J. and Eberts, W. R. (2006) Urban labor
markets, in: R. J. Arnott and D. P. McMillen
(Eds) A Companion to Urban Economics, pp.
389403. Oxford: Blackwell Publishing.
Bartik, T. J. and Smith, V. K. (1987) Urban ame-
nities and public policy, in: E. S. Mills (Ed.)
Handbook of Regional and Urban Economics,
Vol. II, pp. 12071254. Amsterdam: North-
Holland.
Batten, D. F. and Boyce, D. E. (1986) Spatial
interaction, transportation and interregional
commodity flow models, in P. Nijkamp (Ed.)
Handbook of Regional and Urban Economics,
EQUITY IN THE CITY 3219
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
Vol. I, pp. 357406. Amsterdam: North-
Holland.
Bayer, P. and McMillan, R. (2008) Distinguish-
ing racial preferences in the housing market:
theory and evidence, in: A. Baranzini, J.
Ramirez, C. Schaerer and P. Thalmann (Eds)
Hedonic Methods in Housing Markets: Pricing
Environmental Amenities and Segregation, pp.
225244. Geneva: Springer.
Benabou, R. (1993) Working of a city: location,
education, and production, Quarterly Journal
of Economics, 108, pp. 619652.
Berliant, M., Peng, S.-K. and Wang, P. (2006)
Welfare analysis of the number and locations
of local public facilities, Regional Science and
Urban Economics, 36, pp. 207226.
Blomquist, G. C. (2006) Measuring quality of
life, in: R. J. Arnott and D. P. McMillen (Eds)
A Companion to Urban Economics, pp. 479
482. Oxford: Blackwell Publishing.
Brambilla, M. G. and Peluso, E. (2010) A
remark on decomposition of bivariate
inequality indices by attributes by Abul
Naga and Geoffard, Economics Letters, 108,
pp. 493507.
Brueckner, J., Thisse, J.-F. and Zenou, Y. (1999)
Why is central Paris rich and downtown
Detroit poor?, European Economic Review,
43, pp. 91107.
Fleurbaey, M. and Maniquet, F. (2011) A Theory
of Fairness and Social Welfare. New York:
Cambridge University Press.
Gaer, D. van de (1993) Equality of opportunity
and investment in human capital. PhD thesis
Leuven University.
Glaeser, E. L., Resseger, M. G. and Tobio, K.
(2008) Urban inequality. Working Paper No.
14419, NBER, Cambridge, MA.
Gonzales, S. (2009) (Dis)connecting Milan(ese):
deterritorialised urbanism and disempower-
ing politics in globalizing cities, Environment
and Planning A, 41, pp. 3147.
Gyourko, J., Kahn, M. and Tracy, J. (1999) Qual-
ity of life and environmental comparisons, in:
P. C. Cheshire and E. S. Mills (Eds) Handbook
of Regional and Urban Economics, Vol. 3, pp.
14131454. Amsterdam: Elsevier.
Helms, A. C. (2003) Understanding gentrifi-
cation: an empirical analysis of the deter-
minants of urban housing renovation,
Journal of Urban Economics, 54, pp. 474
498.
Kohlhase, J. E. (1991) The impact of toxic waste
sites on housing values, Journal of Urban Eco-
nomics, 30(1), pp. 126.
Lees, L. (2008) Gentrification and social mixing:
towards an inclusive urban renaissance?,
Urban Studies, 45(12), pp. 24492470.
Lucy, W. (1981) Equity and planning for local
services, Journal of the American Planning
Association, 47(4), pp. 447457.
Marans, R. W. (2003) Understanding environ-
mental quality though quality of life studies:
the 2001 DAS and its use of subjective and
objective indicators, Landscape and Urban
Planning, 65, pp. 7383.
Massey, D. and Denton, N. (1996) American Apart-
heid: Segregation and the Making of the Under-
class. Cambridge, MA: Harvard University Press.
Michaels, R. and Smith, V. K. (1990) Market
segmentation and valuing amenities with hedo-
nic methods: the case of hazardous waste sites,
Journal of Urban Economics, 28, pp. 223242.
Michelangeli, A. and Zanardi, A. (2009) Hedo-
nic-based price indexes for the housing
market in Italian cities: theory and estima-
tion, Politica Economica, 2, pp. 109146.
Muth, R. F. (1969) Cities and Housing. Chicago,
IL: Chicago University Press.
Pan Ke Shon, J.-L. (2010) The ambivalent nature
of ethnic segregation in Frances disadvan-
taged neighbourhoods, Urban Studies, 47(8),
pp. 121.
Pratt, J. W. (1964) Risk aversion in the small and
in the large, Econometrica, 32, pp. 122136.
Reardon, S. F. and OSullivan, D. (2004) Mea-
sures of spatial segregation, Sociological Meth-
odology, 34, pp. 121132.
Roback, J. (1982) Wages, rents, and the quality
of life, Journal of Political Economy, 90, pp.
12571278.
Roemer, J. (1998) Equality of Opportunity.
Cambridge, MA: Harvard University Press.
Rosen, S. (1979) Wages-based indexes of urban
quality of life, in: P. M. Mieszhowski and M.
R. Straszheim (Eds) Current Issues in Urban
Economics, pp. 74104. Baltomore, MD:
Johns Hopkins University Press.
Rusk, D. (2003) Cities without Suburbs: A Census
2000 Update. Washington, DC: Woodrow
Wilson Center Press.
Talen, E. (1998) Visualizing fairness, Journal of
the American Planning Association, 64(1), pp.
2238.
3220 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
Tsui, K.-Y. (1995) Multidimensional generaliza-
tions of the relative and absolute inequality
indices: the AtkinsonKolmSen approach,
Journal of Economic Theory, 67, pp. 251265.
Veneri, P. (2013) The identification of sub-
centres in two Italian metropolitan
areas: a functional approach, Cities, 31, pp.
177185.
Weymark, J. A. (2006) The normative approach
to the measurement of multidimensional
inequality, in: F. Farina and E. Savaglio (Eds)
Inequality and Economic Integration, ch. 12.
London: Routledge.
White, M. J. (1983) The measurement of spatial
segregation, The American Journal of Sociol-
ogy, 88, pp. 10081018.
Wong, D. W. S. (1993) Spatial indices of segrega-
tion, Urban Studies, 30, pp. 559572.
Appendix. Derivation of the
Correction Term q in the
Equity-adjusted QoL Index
To derive the correction term q in (5), we
assume a Cobb-Douglas specification for the
evaluation function w:
w(z
i
) =

k
j =1
z
s
j
ij
A1
This specification has several advantages
(1) From the theoretical point of view, the
CobbDouglas form gives a multidimen-
sional index q that generalises the Atkinson
(1970) index and is supported by solid
axiomatics.
14
(2) From the computational point of view, the
CobbDouglas allows one to obtain q by
aggregating k univariate inequality indexes
of the Atkinson type (computed for each
amenity separately) and a residual term
accounting for the interaction among the
distribution of amenities within the city.
(3) The value of q depends on a list of k para-
meters s
j
, with j =1, :: , k, with a clear inter-
pretation: each parameter s
j
indicates the
level of aversion to the unequal distribution
of the amenity z
j
associated with it.
The aggregation property for q and the cali-
bration of parametersthat is the choice of
numerical values for the parametersare
now illustrated. To get q, k unidimensional
indices of the Atkinson type must be
computed
l
j
=
1
z
j
1
n

n
i =1
z
s
j
ij
_ _1
s
j
with j =1, :: , k A2
These indices are interesting in themselves
because they reflect the unequal availability of
the different amenities across city districts.
15
Moreover, one can delve into the contribu-
tion of the distribution of each amenity to the
correction termq, by exploiting the following
Proposition 2: (Abul Naga and Geoffard, 2006)
If w(z
i
) =

k
j =1
z
s
j
ij
, and

k
j =1
s
j
\1; :::, the cor-
rection term q can be decomposed as follows

k
j =1
s
j
_ _
ln q=

k
j =1
s
j
ln l
j
+ ln k A3
where, l
j
are the univariate indices of the
Atkinson class defined in (A2) and k is an inter-
action term equal to
16
k =
n
k1

n
i =1

k
j =1
z
s
j
ij
_ _

k
j =1

n
i =1
z
s
j
ij
_ _ A4
Moving on now to the calibration problem,
notice that equations (A2), (A3) and (A4) depend
on the vector of parameters s= s
1
, :: , s
k
. In
(A2), they enter as parameters of the univariate
indices of the Atkinson type. The higher s
j
, for j
= 1, ., k, the lower aversion to an unequal distri-
bution of the amenity j within the city. A numeri-
cal value must be assigned to s
j
with j = 1, ., k
to carry out the empirical analysis.
17
We first nor-
malise s= s
1
, ., s
k
as follows
s
j
=
1 e
j
k 1
A5
where,

k
j =1
e
j
=1. The normalisation gives a
double advantage: the sum of the k values of s
j
is
1 andmore importantlybecause of the isoe-
lastic form of (A2), the normalisation allows
expressing the Pratt (1964) coefficient of relative
risk (inequality) aversion over the distribution of
EQUITY IN THE CITY 3221
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
amenity j, that is 1 s
j
, as an affine function of
e
j
(1 s
j
coincides with e
j
for k = 2). In this way,
each e
j
represents specific preferences vis-a-vis
inequality. In the benchmark case of an identical
degree of inequality aversion for each amenity,
1 s
j
approaches 1 if the number of amenities
goes to infinite, while it is equal to if k = 2.
Notice that and 1 are the two parameters usu-
ally chosen by the Luxembourg Income Study
for the Atkinson index.
18
In our analysis, the
degree of inequality aversion is allowed to vary
with amenities. Since implicit prices truly cap-
ture the assessment of each amenity by the citys
representative citizen, we pragmatically assume
that the higher the contribution of an amenity in
determining the QoL value in the city, the more
intense is the aversion for its uneven distribution
within the city. This is enshrined in the following
assumption
Assumption: The relative inequality aversion coef-
ficient e
j
, with j = 1, ., k, is set to be equal to
e
j
=
p
j
z
j

k
l =1
p
l
z
l
A6
Each parameter e
j
is set to be equal to the ratio
between the estimated value of the average
quantity of the amenity j and the value of all
amenities which are on average available in the
city. In other terms, e
j
corresponds to the relative
contribution of amenity j to the overall QoL
index.
All in all, the methodology can be summarised
in the following steps
(1) The available amenities z
ij
are computed
using expression (2).
(2) The implicit prices of the k available ame-
nities are estimated through the hedonic
regression (7).
These two steps are sufficient to compute the
QoL index adjusted for the availability of the
amenities (4). To compute the equity-adjusted
QoL index (5) the procedure also includes the
following three steps
(1) The coefficients e
j
, with j=1,..,k, are com-
puted from equation (A6).
(2) Vectors s= s
1
, . . . , s
k
and l =
l
1
, . . . , l
k
are computed using expres-
sions (A5) and (A2) respectively.
(3) The correction term q is obtained from
equation (A3) and QoL
q
(Z; W) is finally
determined.
3222 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
Table A1. List of neighbourhoods in Milan
Zone Neighbourhood name Inhabitants Average income (e)S QoL (e)
Centre
1 Scala, Manzoni, Vittorio Emanuele, S. Babila 3 265 60 257 263 453
2 Brera, Duomo, Cordusio, Torino 7 712 61 387 146 553
3 Missori, Italia, Vetra, SantEufemia 6 134 45 529 161 974
4 Diaz, Fontana, Europa 4 448 51 638 179 833
5 Cadorna, Monti, Boccaccio 6 493 57 335 130 176
6 Castello, Melzi dEril, Sarpi 24 478 30 850 74 775
7 Turati, Moscova, Repubblica 7 975 41 599 138 516
8 Venezia, Majno, Monforte 2 658 76 556 194 809
9 Mascagni, Porta Vittoria, Porta Romana 16 355 49 959 99 891
10 Porta Ticinese, Porta Genova, Magenta 18 958 33 647 103 477
Mid centre
11 Cenisio, Procaccini, Firenze 26 088 20 024 52.749
12 Fiera, Giulio Cesare, Sempione 19 756 28 289 77.183
13 Amendola, Monte Rosa, Buonarroti 17 566 32 762 74.483
14 Pagano, Monti, Wagner 8 448 44 901 107.130
15 Piemonte, Washington, Cimarosa 32 651 28 928 90.539
16 Solari, Napoli, Savona 10 757 23 921 101.283
17 Naviglio Grande, Argelati, San Gottardo 13 353 18 715 68.724
18 Tabacchi, Sarfatti, Crema 31 802 21 618 65.100
19 Libia, XXII Marzo, Indipendenza 52 173 22 264 56.482
20 Regina Giovanna, Pisacane, Castel Morrone 22 410 26 847 90.670
21 Abruzzi, Eustachi, Plinio 22 906 26 321 106.415
22 Stazione Centrale, Gioia, Zara 50 087 22 457 82.059
Outlying
23 Musocco, Varesina, Certosa 36 200 14 740 35 251
24 Bovisa, Bausan, Imbonati 32 519 14 595 42 884
25 Largo Boccioni, Aldini, Lopez 15 514 11 887 35 848
26 Bovisasca, Affori, P. Rossi 43 240 14 335 40 679
27 Niguarda, Ornato 25 397 14 411 51 490
28 Fulvio Testi, Bicocca, Ca Granda 26 694 15 885 62 749
29 Monza, Precotto, Gorla 30 814 15 458 50 086
30 Zara, Istria, Murat 14 361 16 034 52 883
31 Loreto, Turro, Padova 51 457 16 813 69 693
32 Parco Lambro, Feltre, Udine 56 571 14 993 47 346
33 Aspromonte, Porpora, Teodosio 26 259 19 721 67 789
34 Leonardo da Vinci, Gorini 19 303 20 247 65 149
35 Lambrate, Rubattino, Folli 6 768 16 518 42 878
36 Argonne, Viale Corsica 27 318 20 777 59 496
37 Forlanini, Mecenate, Rogoredo 32 255 14 540 34 697
38 Ortomercato, Molise, Piranesi 12 071 18 783 50 896
39 Boncompagni, Toffetti, Bacchiglione 11 933 17 235 41 159
40 Omero, Gabriele Rosa, Brenta 20 996 13 447 34 283
41 Ronchetto, Chiaravalle, Ripamonti 12 158 17 404 37 389
42 Montegani, Cermenate, Vigentino 45 664 16 991 49 032
(continued)
EQUITY IN THE CITY 3223
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from
Table A1. (Continued)
Zone Neighbourhood name Inhabitants Average income (e)S QoL (e)
43 Barona, Famagosta, Faenza 46 458 14 761 50 401
44 San Cristoforo, Ronchetto, Ludovico il Moro 22 260 12 360 55 367
45 Giambellino, Tirana, Frattini 27 723 16 396 45 903
46 Siena, Tripoli, Brasilia 42 150 19 313 66 706
47 Lorenteggio, Inganni, Bisceglie 41 657 14 848 50 972
48 Novara, San Carlo, Amati 9 704 14 729 47 383
49 Segesta, Capecelatro, Aretusa 29 543 16 204 46 154
50 Ippodromo, Caprilli, Monte Stella 5 070 27 664 91 340
51 Cagnola, Achille, Papa, Tiro Segno 5 826 19 907 52 780
Suburbs
52 Baggio, Quinto Romano, Quarto Cagnino 45 757 14 107 36 115
53 Gallaratese, Lampugnano, Figino 44 020 18 652 66 154
54 Missaglia, Chiesa Rossa, Gratosoglio 20 624 12 595 40 792
55 Quarto Oggiaro, Roserio, Amoretti 17 894 11 791 30 096
Table A2. Description of variables
Variable Definition Source
Market_p 1 if the housing value is a market price (ref.) OMI
Offer_p 1 if the housing value is an offer price OMI
Estimated_p 1 if the housing value estimated by OMI OMI
Private characteristics
ln(size) logarithm of the total floor surface area OMI
Second bathroom 1 if the unit has a second bathroom, or more OMI
Third bathroom 1 if the unit has a third bathroom, or more OMI
To be renewed 1 if the unit needs to be renewed OMI
Heating 1 if the unit has gas central heating OMI
2nd floor (or higher) 1 if the unit is on second floor, or higher OMI
Low cost 1 if the unit is in a low-cost building (ref.) OMI
Medium cost 1 if the unit is in a medium-cost building OMI
Luxury 1 if the unit is in a luxury building OMI
Parking 1 if the unit has at least one parking space OMI
Elevator 1 if the unit is in a building with an elevator OMI
ln(age) logarithm of the age of building OMI
Distance distance of the unit to the centre Authors computation
City centre 1 if the unit is one of the zones in the centre OMI
Amenities
Transport number of metro stations Milan Transport Agency
Proximity_university proximity to the nearest university Authors computation
Green percentage of public green areas Milan Municipality
Cultural cultural places
a
per 10 000 inhabitants Yellow Pages
Shopping_facilities commercial facilities
b
per 10 000 inhabitants Authors computation
Ethnic Italian/foreign Milan Municipality
a
Cinemas, theatres, museums, art galleries, academies of music and libraries.
b
Supermarkets, discount stores and malls.
3224 MARCO BRAMBILLA ET AL.
at CEPT University on July 28, 2014 usj.sagepub.com Downloaded from