Towards a Comparative Science of Cities: Using Mobile Traffic Records in New York, London, and Hong Kong - Computational Approaches for Urban Environments - page 358

city center. In New York, the organization is polycentric, with one center in the
middle part of Manhattan and another one in Queens. Finally, in Hong Kong we
observe one big center of activity divided up between Kowloon and the northern
part of Hong Kong Island and secondary centers in the newly developed zones of the
New Territories. Low activity zones correspond to the limits with the mountainous
areas.
Lorentz curves depicting the variation of cumulative total request activities with
the cumulative areas of coverage are presented on Fig. 15.2 d. These curves, typically
used in economy and ecology to describe inequality in wealth or size Lorenz
( 1905 ), describe here unequal repartition of request activity in space. Hong Kong
is obviously the most inhomogeneous city (with less than 2 % of request activity
in the 50 % less active pixels and 64 % of activity concentrated in the top 10 %
most active pixels), followed by London (around 9 % of activity in the 50 % less
active pixels and 49 % of activity in the top 10 % most active pixels) and New York
(around 9 % of activity in the 50 % less active pixels and 38 % of activity in the top
10 % most active pixels).
A commonly used quantitative measure of inequality, the Gini coefficient, can be
defined from a Lorentz graph as the area between the bisector line (corresponding
to a uniform repartition) and the Lorentz curve, normalized by the area between the
bisector line and the x axis (corresponding to the most inhomogeneous case where
all activity is concentrated on one pixel) (Gini 1912 ). The Gini coefficients of the
request Lorentz curve of Fig. 15.2 d as well as those corresponding to other types
of activity are reported in Table 15.2 . Interestingly, the Gini coefficient depends
only slightly on the activity type and strongly on the city. The measure of spatial
inhomogeneity could thus be done on any type of activity. Again, we find here that
the most inhomogeneous city is Hong Kong, followed by London and then New
Yo r k .
Although this first analysis suffers from some limitations - such as the mismatch
between the area covered by our dataset in London and the area within the official
boundaries of Greater London - it already provides good insights on the way people
interact with their cities. Maps of mobile phone activities could steadily become a
complementary tool to more classical maps of population or employment densities
obtained through extensive surveys and help urban planners make decisions based
on accurate population repartition.

Gini coefficients For each activity type , the Gini coefficient G Table 15.2 2 Œ0; 1 measures

the inhomogeneity of the spatial repartition of the three-month aggregated activity. The higher the
coefficient, the more unequal the spatial repartition is
City G DL Data G UL Data G Request G Calls G SMS

Greater London 0.608 0.640 0.649 0.618 0.606

New York 0.546 0.555 0.576 0.549 0.523

Hong Kong 0.768 0.765 0.784 0.802 0.781