ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Cluster and Network Analysis Techniques in a Multi-Stage Methodology for Business Research

Emanuela Todeva, David Knoke, Donka Keskinova

Building: Law Building
Room: Breakout 2 - Law Building, Room 026
Date: 2012-07-10 01:30 PM – 03:00 PM
Last modified: 2011-12-19


The use of large business databases containing longitudinal secondary data and heterogeneous indicatos on industries, firms and their activities and performance has posed a range of methodological questions regarding the value and validity of secondary data, the categorisation and representation of data, the intrinsic relationships between different categories of data and the visualisation of these relationships as well as the visualisation of the data itself. We have embarked upon the development of a multi-stage methodology that combines multiple cluster analysis techniques with correspondence analysis and network analysis techniques in order to reveal structural relationships in industries and strategic groups of firms in the global market space. In this paper we present one of the undertaken projects that has used our multi-stage cluster methodology.

For this analysis we have developed an original database of firms that consists of the 150 largest corporations identified by Fortune 500, Fortune 1000 and Global 500 lists from 1989-2008 as operating in the global information sector (GIS). Data on the primary and multiple secondary industry SIC codes for each firm in 1998 and 2008 were recorded, along with other performance indications and context data on the country of origine and counties of operation. We analyse the segmentation and the industry structure of the GIS, using hierarchical cluster analysis and correspondence analysis. We calculate proximity measures to estimate the co-location of firms in ‘product fields’. On the basis of these proximity measures we identify 23 clusters of firms that participate in strategic groups (clusters) accoring to their four-digit American SIC (Standard Industrial Classification) codes. The cluster co-location of firms is interpreted as a structural relationship that associates global firms in strategic groups with their direct competitors.

The subsequent correspondence analysis reveals a very strong relationship between cluster membership and the Primary SIC code that describes the core capabilities of the GIS firms. It also reveals that the 23 clusters occupy 6 distinctive structural spaces that correspond with more traditional sectors in the economy. Another correspondence analysis between cluster membership and country of origin reveals clear patterns of clusters strongly associated with a particular home market - Europe, Asia, or NAFTA, and other group of clusters that share expertise between two of the three Geo-economic centres (Fig. 3).

The main methodological steps in our approach are: first, building the business database, second identifying similarity measures and applying to firm database – to categorise firms in strategic groups (clusters); third, applying correspondence analysis to verify clusters and to evaluate inter-cluster relationships, to test the robustness of our cluster formations, and to reveal new cluster attributes that could explain cluster dynamics; and fourth, applying one-mode and two-mode network analysis to map inter-firm, inter-cluster and firm-attribute relationships.

This multi-stage cluster methodology has been developed to address issues that go beyond a multi-method or mixed method research. It is based on the utilisation of existing depositories with business data, and enables to undertake a structural analysis revealing patterns, ontological categories. relational attributes and structural relationships in these datasets through primary, secondary and tertiary analysis of data.