Modular network comparison with application to labour flow networks

Thesis

Evolutionary economic geography emphasises the role of locally embedded skills and knowledge in regional economic development processes. A region grows into new economic activities by combining locally available skills. This is typically modelled as a branching process on an inter-industry labour flow network, where a region’s probability of entering an industry is a function of the related industries available. In this thesis, we develop new methodologies that support the analysis of these labour flow network models.

In the first part of this thesis, we develop new metrics to compare the modular structure of different networks. The motivating application is the comparison of different countries’ inter-industry labour flow networks. More specifically, we present a novel approach for the global modular comparison of two node-aligned networks. The measure quantifies the difference in the quality of the partition of two graphs by projecting the community structure of each graph onto the other. The technique is advantageous as it captures the influence of the underlying network structure and applies to a broad range of networks as it is agnostic to the quality function adopted.

Building on this framework, we also present a fine-grained modular comparison technique, which compares the modular structure of a single community across two networks. The measure compares the expected escape probability of a random walker, initially starting in the community, across the two networks. The measure is advantageous as it captures the impact of edge density and connectivity on the modular structure of the community.

In the second part of the thesis, we investigate two economic questions. First, we investigate whether the presence of related multinational enterprises (MNE) results in knowledge spillovers that foster new related domestic industries in Irish regions. We propose a new dynamic-based cohesion measure for an inter-industry labour flow network that captures the impact of higher-order linkages between an industry and the industrial portfolio of a region. We find that industries containing both domestic and MNE firms enhance the entry and survival of related domestic industries within a region, while industries dominated by MNEs have the opposite effect.

Finally, we consider the indirect impact of an affirmative action policy on inter-industry gendered labour mobility patterns in South Africa. By harnessing a regression discontinuity design with an entropy-based centrality measure, we investigate how the policy has impacted the sectoral diversity of new employees in firms. We find that the policy widened the sectoral diversity of female workers, particularly the related variety of skilled female workers. However, unlike results in the literature, this did not translate into any significant impact on firms’ performance.

Authors: Landman, MS

• DOI: 10.5287/ora-erzqezop8
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: labour flow networks; network comparison; economic geography; knowledge flows
• Subjects: Applied mathematics; Social sciences--Network analysis