To start with a triviality - as any real-world
topic, development in a developing country has an uncountable
number of important and different aspects – what makes it
difficult to gather and assess sensible information on it, and,
especially, to draw sound conclusions from these. This task starts
with a definition or a framework to confine the term "development"
in order to pose more concrete questions. In a second step, methods
to assess the development and to report and give interpretations
of the results have to be chosen. There is no unique way to achieve
this. This is mirrored in the presence of several approaches to
define, to measure and to assess development on different levels
of aggregation and formalisation and in different topical frames.
This project deals with development on the household level, focusing
on socio-economic and infrastructural characteristics as reported
in the household expenditure survey data of India we base our
research on. Our approach is based on the idea that groups can
be found or defined on having similar expenditure characteristics,
both generally and in terms of their energy use pattern. The evolution
of household energy use can then be modelled by the change in
size, even disappearance of existing or emergence of new groups
with their distinct energy use pattern.
To identify potential groups, we employ several statistical tools
designed to reveal nontrivial intrinsic structure in data sets.
These are cluster analysis and dimension reduction techniques
such as factor or principal component analysis. In addition, we
build groups of households by combination of the different levels
of several categorical variables. The resulting groups are then
assessed on grounds of their energy use patterns and the evolution
thereof over the years.
A potential difficulty with cluster analysis and dimension reduction
techniques is the necessity of a considerable amount of assumptions
and subjective choices that have to be made by the researcher.
This bears the danger that the resulting groups do not reflect
any intrinsic structure of the data but are only generated by
these choices. On the other side, the potential drawback of groups
defined by levels of categorical variables is their triviality.
In contrast to the other methods, however, such groups are more
transparent and independent of implicit assumptions.
These reservations are duly kept in mind while assessing the
different methods to identify intrinsic groups, different assumptions
and the results they produce, their adequacy to capture development
and the picture they draw of the development of the whole society.
