Why would one want to model the land use and development outcomes of an entire city? How would anyone attempt to simulate such a complex process? What’s in it for policy makers, urban planners and property developers?
‘Machine learning’ is now capable of processing huge data sets of cross correlated information; to observe urban development patterns dynamically over time. Investment in a new park, rezoning, street trees, new infrastructure such as a tram, or simply redirected market interest, can change property demand in an area, tempered by the rate of housing supply. Property prices as a result go either up, or on occasion can go down. These are overlayed by more complex and metropolitan wide patterns such as gentrification waves or financial and policy inputs that can impact property demand over entire metropolitan regions.
PointData have developed a set of algorithms to tackle such urban challenges. Version 1.0 of its new software (Urban Policy Model) can be used by State Governments, Councils, utility companies and even the development industry to predict urban outcomes. It can determine location unique supply- demand price curves at a suburban level while drawing on tens and even hundreds of thousands of parcel data sets simultaneously. House and land costs, development patterns and housing market choices can also be derived for individual properties. This holistic modelling is constantly interacting in a bottom up, top down way to provide each individual parcel with a more accurate and predictive future property value, that can be aggregated up to represent a set of metropolitan indicators.
What will the software do?
The software can combine with Geographical Information Systems GIS) to map a development area, zone extent or entire council area to visualise scenarios, which can include: investment into new social and physical infrastructure including hospitals, open space, access points, as well as property uplift and value capture analysis from public transport investment.
Infraplan applied the model to a new retail offer (above) to test zoning changes around a shopping centre located in the middle ring of Adelaide. Drawing on the strong Italian local heritage, it proposed strengthening the specialty food business, inspired by overseas models such as Eataly.
The investment was predicted to deliver land value uplift, market attractiveness and consequently an increase in the local population catchment within a 5 minute walk of the centre. Zoning tweaks were then tested to drive the optimum urban density outcome. This complex push-pull policy analysis was tempered by wider market forces that reduced the number of dwellings (yield) based on land values/market preferences to certain housing types. The resultant metrics included 2031 projections for population, parking through to traffic and parking demand and council rates.
Using Artificial intelligence to find patterns and test planning and financial policies
Machine learning presented itself as a better way of interpreting complex patterns at both a city-wide and detailed property level, essentially supercharging the process. This enabled the architecture of our algorithm to evolve to include numerous ‘training’ data sets (to teach our algorithm metropolitan wide patterns) and error metrics (even more algorithms!) to test the outcomes of the model against ‘on the ground’ market observations.
Financial levers – The Urban Policy Model (UPM) can measure the impact of government taxes and levies such as infrastructure changes and stamp duty breaks market changes affecting feasibility including interest rates, supply/demand, real estate price growth and employment and building industry changes, such as costs and inflation, margins, commissions and development fees.
Planning Policy levers – The UPM has since evolved to test policy scenarios and analyse zoning changes, such as site and area constraints, density targets, floor ratios, height limits, minimum frontages and open space requirements. It can also test the redevelopment of low density areas, as well as opportunities along transit corridors and in centres as well as complex calculations such as site amalgamations.
Property uplift and value capture testing
Unlike some LUTI models based on regression analysis or gravity models that predict land use responses to accessibility improvements from transport investment the UPM works at a micro-simulation level getting a precise land use response. Property uplift falls away the further the property is located from the transport, park or other infrastructure investment. Moreover, every property is capitalised and unique in shape and size meaning that it will respond (property uplift) to investment in a completely different way to say an adjacent property. In other words property uplift and urban density responses will not be homogeneous along say a transit corridor. Traditional approaches tend to grossly overstate property uplift/value capture outcomes, especially in moderately growing areas.
PointData is currently inviting state and local governments, and utility companies, to express their interest in building an Urban Policy Model for their metropolitan jurisdiction. PointData has now employed product development systems based on an Agile/LEAN business model, which many Silicon Valley tech sector companies adopt themselves.
PointData is also inviting sophisticated investors and industries partners to participate in this exciting venture. Please contact us for further information.