Spatial computable general equilibrium (SCGE) models provide a comprehensive and economically coherent framework for analysing public investment or policy options in metropolitan areas and regions. The Victoria University (VU) Cities SCGE framework is being applied to assess proposed transport infrastructure investments and land use planning strategies in the Australian states of New South Wales (NSW) and Victoria. In those applications, VU Cities models are coupled with strategic transport models to handle mode choice and congestion effects. In this paper, we illustrate the application of the VU Cities--NSW model with a counter-factual analysis of an outer suburban rail loop in Western Sydney that connects to the existing rail network. We base our analysis on estimates of: (i) the annualised cost of having provided this infrastructure and (ii) the direct effects on generalised travel costs between all relevant origins and destinations. We first elucidate the long-run impacts on the spatial distribution of housing and employment. Our main focus, however, is on quantifying and explaining the overall costs and benefits and examining their sensitivity to several key modelling assumptions. As the financial costs are outweighed by gains in productivity, amenity and commuting time savings, more residents and workers are attracted to the city. Population gains are reinforced by positive externalities of residential density but also push land rents up and wage rates down. We show that these latter effects are sensitive to assumptions about the strength of both positive externalities and the operation of density-suppressing land use regulation. However, on a per capita basis, changes are much smaller and also less sensitive to these assumptions.