Enhancing land-use inputs to strategic transport models through advanced econometrics: CUBE Land Best Practice Review: Literature and User Survey

The overall aim of this project was to improve household and business land use activity inputs to transport modelling provided by a strategic land use model. The particular focus has been placed on the proprietary software CUBE Land being implemented for Perth and Peel by the Department of Planning, Lands and Heritage (DPLH) with support from the Western Australian Planning Commission (WAPC), DoT and MRWA. The project objectives were to:
• Provide an enhanced ‘under the hood’ understanding of advanced econometrics as applied in practice in spatial econometric land use models with particular emphasis on CUBE Land as applied in Perth and Peel, WA; and
Capture and share the learning from a review of the application and use of the bid-rent model structure and parameter estimation elsewhere in the world, in order to provide recommendations for enhancing the implementation of the Perth and Peel CUBE Land model, given available data and software limitations.
The project comprised two parts:
• Delivering an advanced econometric training course (Milestones 1 and 2); and
• Conducting a best practice international review of the bid-rent model (Milestone 3).
The PATREC team encourages the DPLH to continue the advancement of the CUBE Land model for Perth and Peel, acknowledging the progress made at a rapid pace and the potential for further innovations, stemmed from the group’s commitment and ability to generate and communicate new knowledge. As always, deadlines, asynchronous interactions with the developer, increasing costs and in-kind contributions, fear of scepticism or potential negative feedback from other stakeholders or cross-agencies, and finally the risk of unforeseen complications, may act as hurdles to embracing CUBE Land as a tool to assist planning for WA. Yet, behavioural models, explaining choices made by households and businesses, are undoubtedly superior to sketch models relying heavily on simplify assumptions and the models aiming only for predictive power. Nevertheless, the model needs further calibration and validation. We thus support validation through a number of processes: sensitivity analysis and scenario development, testing various models with distinct inputs, and cross-validation with the outputs from the land use projection system developed parallelly within DPLH such as the Dwelling and Employment Distribution System (DaEDS).
We present below a number of aspects deemed appropriate for consideration by the DPLH CUBE Land team. These actions will significantly improve the performance of CUBE Land, which is expected to provide meaningful forecasts of residential and non-residential use for urban and transport planning in a much shorter timeframe than other jurisdictions:
New functions for bid-rent and cost should be used; Logsum accessibility improvements; Clustering of agents should consider homogeneity of location behaviour and its determinants; Hard constraints should be further examined; Outliers which may have deleterious effects in the rent and cost models should be removed; Start with the estimation of the bid-rent (demand) and cost (supply) models for non-residential properties; providing essential detailed information on the data input quality expectations; Sensitivity analysis as a priority