Flowbca: A Flow-Based Cluster Algorithm in Stata

Abstract

In this article, we introduce the Stata implementation of a flow-based cluster algorithm, flowbca, written in Mata. The main purpose of flowbca is to identify clusters based on relational data of flows. We illustrate the command by providing multiple examples of applications from the research fields of economic geography, industrial input-output analysis, and social network analysis.

Agglomeration Economies and the Urban Wage Premium in Australia

“…The regional aggregations are defined by applying a flow-based cluster algorithm, entitled flowbca, introduced by Meekes and Hassink (2018). 7 The continuum of regional aggregations contains sets of spatial units that range from 2,164 to 10 unique spatial units, which are aggregated one-by-one using a hierarchical clustering algorithm.…”

Agglomeration Economies and the Urban Wage Premium in Australia

“…The regional aggregations are defined by applying a flow-based cluster algorithm, entitled flowbca, introduced by Meekes and Hassink (2018). 7 The continuum of regional aggregations contains sets of spatial units that range from 2,164 to 10 unique spatial units, which are aggregated one-by-one using a hierarchical clustering algorithm.…”

Endogenous Local Labour Markets, Regional Aggregation and Agglomeration Economies

“…We use flowbca, discussed by Meekes and Hassink (2018), which is an implementation of a flow-based agglomerative hierarchical cluster algorithm that is able to define LLMs by clustering disaggregated spatial units into aggregated spatial units. 10 We define LLMs for different subgroups of workers at various levels of regional aggregation.…”

“…We follow the literature by focusing on differences in workers' LLM through differences in workers' commuting flows from residence to workplace (e.g., see Farmer and Fotheringham (2011); Brezzi et al (2012); Duranton (2015); Foote et al (2017)). 3 We use a flow-based cluster algorithm, flowbca, to define multiple sets of LLMs that vary in spatial unit sizes (Meekes and Hassink, 2018). 4 The main input for flowbca is relational data on commuting flows that the algorithm uses to iteratively group two spatial units into one.…”