International Journal on Criminology Volume 3, Number 2, Fall 2015 | Page 97
Burglaries in France
methods. These are used to synthesize the available information, thus making it simpler
to read and interpret. This study makes use of two multivariate analysis approaches.
The first is multiple correspondence analysis (MCA), which is used to
synthesize all of the relevant factors and values into two composite indicators. It
provides a graphical approach that aims to represent variables in a diagram and to
measure their “proximity” or “likeness.” This technique is not intended to create a
priori categories; nevertheless, the graphical representation provides an illustration of
the groups created during the typology (see below).
The second is classification, which facilitates the creation of a typology of
individuals in the statistical sense of the term (here, households), thus allowing them
to be grouped into homogeneous sets according to certain characteristics. 4 We used
these two techniques in a complementary manner and for exploratory purposes on
profiles of households in mainland France, relying on the characteristics presented
above.
The synthesis of information through MCA and the creation of a typology of
households were subsequently used to estimate the rates and the differences in the
rates of victimhood with regard to burglaries between different groups. As a result, it
was possible measure the proportion of victims within each group and determine the
profiles most at risk among the groups identified.
Results
Creation of the Typology
Here we jointly analyze the graphic results of the MCA and the creation of the
typology resulting from the classification for each factor (sociodemographic,
accommodation, neighborhood). It is worth recalling that MCA is not
supposed to establish a priori groups, though the illustration that it offers is in line
with the groups created by the classification. Readings of these results are therefore
complementary.
The first MCA was carried out for households’ socioeconomic conditions.
Figure 1 allows the distance between the values of the characteristics to be visualized
via a two-dimensional diagram, 5 and allows associations between them to be identified.
4
Readers interested in a fuller introduction to these methods may wish to consult Dehon, Droesbeke,
and Vermandele (2008) or Saporta (2006).
5
Two “axes” were defined for each of the three MCAs. Each axis is a combination of all the factors of
origin and each factor makes its own contribution in each axis. Thus, the position of a value in the
diagram depends on its weight on the first axis and on the second axis. For example, the value “unemployed”
has a high and positive weight on the vertical axis and a very low weight on the horizontal
axis. The reading of the position of each value taken individually allows the axes to be characterized,
and the joint reading of the values allows the link that may exist between them to be demonstrated.
The values associated with the axes (dimensions) in the graphs represent the representational capacity
of each axis in the total dispersion of the values. In the case of figure 1, the two axes used allow around
18% (10.67 + 7.45) of the total dispersion of the values to be explained.
92