• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site
  • HSE University
  • Publications of HSE
  • Articles
  • Паттерн-анализ и кластеризация в исследовании государственной состоятельности: "адаптивная оптика" для политической науки


Паттерн-анализ и кластеризация в исследовании государственной состоятельности: "адаптивная оптика" для политической науки

Политическая наука. 2019. № 3. С. 112-139.

The central focus of this paper is a methodological one. Using the set of indicators
of state capacity, we demonstrate a specific strategy for identifying sustainable
structures in multidimensional data sets that reflect complex and ambiguous concepts of
political science. A key feature of this strategy is the application of related, but significantly
different technically, multidimensional methods – cluster and pattern analyses.
We use hierarchical clustering with various combinations of metrics and amalgamation rules, as well as ordinal-invariant pattern-clustering. Properties of pattern analysis as a
method for studying multidimensional data are shown for the first time (to the best of
our knowledge) in the political science literature. Since clustering has been actively
used in political science for a long time, pattern analysis is still practically not adopted
in our science. This situation requires correction, since pattern-analysis has some important
and in many ways unique capabilities.
It was shown that the combination of pattern and cluster analyses makes it possible
to identify consistent structures that have a clear interpretation in terms of political
science. Thus, in the course of our study, several types of state capacity were identified
(although this task was rather illustrative for us).
We use a set of empirical indicators of state capacity: the share of military
spending in GDP, the share of military personnel in the total population, the share of
tax revenues in GDP, the total rate of homicides and victims of internal conflicts, and
the quality of government institutions. Data for more than 150 countries are taken for
1996, 2005 and 2015. Stable combinations of the values of these indicators, identified
simultaneously via pattern and cluster analyses, form the structures of state capacity.
In conclusion, we show the most promising directions for the development of
the methodology described in this paper. One of the most important is the analysis of
the dynamics of countries within the pattern-cluster structures of state capacity.