The purpose of this tool was to develop an open-source software tool for identifying urban employment clusters, that will be accessible to anyone with a suitably-appointed computer and internet access. The tool enables governments and researchers could examine spatial employment clustering in metropolitan regions in much finer spatial units than are currently supplied by the Australian Bureau of Statistics.
This project responds to a consensus among local policy makers, that Melbourne needs to adopt a multi-nodal metropolitan planning strategy in order to foster local economic development and reduce commuting. For decades, metropolitan planning strategies have sought to promote non-CBD centres in Melbourne. The tool further responds to a consensus among economic development planners that ABS data is insufficient to identify local urban clusters for analysis. We wish to understand whether spatial policies aimed at cluster development have actually resulted in employment clusters. This tool moves us toward examining those policies by providing a framework to identify whether and where local employment clusters have formed.
Our tool provides an analytical process by which researchers can identify spatial clusters of industry in the Melbourne Metropolitan Region (MMR). In turn, these clusters can then be used to address issues of concern to urban policy making in the region – issues that researchers and analysts have previously struggled to address due to a lack of fine spatial data on employment location. Sample policy questions are:
- What industry clusters exist in the Northwest Corridor? Are they located inside or outside the planned cluster areas? Have sector-specific clusters emerged, either in the activity centres or outside of them (e.g., education clusters, justice service clusters, biomedical and biotechnology clusters, high-technology clusters)?
- Where do the metropolitan area’s workers live, how far do they travel to work, and how accessible are job opportunities? How are their travel choices and trends related to local transport provision? Have clusters made jobs more accessible to workers in the Northwest Corridor?
- What spatial effects can be identified in workforce travel behaviour and accessibility in Activity Centres in the Northwest Corridor – specifically for Footscray, Broadmeadows and Melton? Do travel patterns in these areas differ from other non-CAA places?
- How do travel patterns, accessibility of jobs, and job location differ between those for key service workers (nurses, teachers, etc., who are more likely to be disadvantaged) and the mainstream population? Have CAAs addressed the needs of these groups?
- What is a typical commute and accessibility profile of a cluster-based employee, versus an employee that travels to a dispersed workplace? Which types of clusters attract the most workers and jobs? How far is the reach of the clusters – i.e., how far does the commute sheds (areas from which clusters draw commuters), extend for the clusters in the Northwest Corridor?
To build the tool, we have used the open-source Cran R spatial analysis packages. After the user specifies an industry of interest, the tool splits Census Destination Zones (DZNs) with jobs in the industry of interest into smaller polygons based on land use data, and attributes Census Journey to Work (JTW) job destinations to each smaller polygon. Then, a modified Ward’s algorithm clusters the non-contiguous small polygons using spatial and non-spatial attributes. The tool also allows the user to specify whether clustering should be limited by local or regional scale.
Who are the Users and the Testers?
The tool was initially developed and tested with a working groups comprising of representatives from academia, state, and local government. The main policy question raised by the users was the need to efficiently identify where employment clusters are. The tool has been designed to respond to that need. The tools allow the users to identify clusters of employment, disaggregated by industry type. We foresee many applications of the tool including spatial land-use planning, transport investment decision-making, and activity centres policy analysis, especially concerning the commuting practices and patterns of those working in them.
Testing Methods and Findings
The tool has first been tested within the R programming environment. This processesnabled the analytics of the tool to be tsted. After this testing the tool was programmed into the AURIN Portal. This process has required rigerous technical testing to ensure compatibility and stability of the tool. Making the tool available within the AURIN portal means that the data is secure and the tool is available to researchers and planners (who have access to the portal).
Findings and Lessons Learned
1 Regional versus Local Scale
Figures 1 through 3 illustrate example outputs based on cluster configurations with 5, 10, and 20 km thresholds for Motor Vehicle and Motor Vehicle Part Manufacturing defined by the Australian New Zealand Standard Information Codes (ANZSIC) 2310, 2311,2312, 2313, 2319.
The labels in the map are of the corresponding local government area (LGA), whose boundaries are shown in light grey. Variation in distance threshold specifications has a large impact on the number and types of clusters formed. A small distance threshold generates more clusters, and thus a smaller average number of jobs per cluster, since the algorithm distributes the same number of jobs over more clusters. When we set our distance threshold to one kilometre, the algorithm generated 145 clusters in total, with 119 of these containing more than 50 jobs. At a distance threshold of five kilometres, the number of clusters is reduced drastically to 23
Image 2: Cluster Configuration, ANZSIC Code 2310, 10km Threshold, Spatial Weight = 0.90
Image 3: Cluster Configuration, ANZSIC Code 2310, 20km Threshold, Spatial Weight = 0.90
This project was difficult because current Australian Census geographies do not provide a way to analyse sub-metropolitan clustering. This was problematic for us because we were seeking to analyse metropolitan clustering policies in attempt to understand whether they are resulting in the intended effects. To overcome this lack of consistent data a series of scripts were developed to integrate data from the Victorian Government Planning Scheme, Valuations Data in combination with the ABS Census and Journety to work data boundaries.