American Association of Geographers 2018

This is the presentation I gave at the AAG 2018 in New Orleans. It discusses ongoing work on high resolution demographic projections.

Transcript

1. Nik Lomax
   Andrew P Smith
   AAG New Orleans
   Innovations in Urban Analytics, 12 April 2018
   High resolution demographic
   projections for infrastructure
   planning
   

   View Slide

2. • The UK Infrastructure Transitions Research Consortium
   (ITRC) is a collaboration of seven universities and over
   50 partners from infrastructure policy and practice.
   • ITRC is helping governments, utility providers, designers,
   investors and insurers by developing new ways to
   evaluate the performance and impact of long-term plans
   and policy for infrastructure service provision in an
   uncertain future.
   • Supported by an EPSRC Programme Grant.
   

   View Slide

3. Demographics for demand
   models
   Demographic projections
   Demand
   modelling
   

   View Slide

4. Essential for planning the
   delivery of services and
   the allocation of resources
   to sub-national areas
   2014 sub-national population projections by ONS
   projections of diversity to 2061 from NewEthpop
   Or the assessment of population change
   for measuring demographic diversity and
   social equality
   Demographic projections
   

   View Slide

5. Demographic projections
   Projections are needed at high
   resolution for the MISTRAL project
   Crown copyright © 2017 Ordinance Survey
   2014 sub-national population projections by ONS
   But Spatial resolution is
   usually limited to larger
   administrative areas
   

   View Slide

6. Two projection approaches
   • Static sequential microsimulation constrained to
   official projections
   • Dynamic microsimulation which borrows strength
   from survey and census data as well as supply data
   from new build and housing stock information
   • The static model is delivering results, the dynamic
   model is work in progress
   

   View Slide

7. Data inputs
   • 2011 Census
   • Standard tables – near complete enumeration
   • Microdata – detailed attribute cross-tabulations
   • Survey data to add attributes/ calculate transition rates
   • British Household Panel Survey
   • Understanding Society
   • Ordinance Survey Mastermap & Address Point data
   • ONS population projection constraints
   • DCLG (1) household projections and (2) housing additions
   

   View Slide

8. Modelling approach
   

   View Slide

9. Population at
   MSOA scale
   Classified
   residential
   buildings
   Residential
   household size
   and type
   Demand models
   Model component integration
   Households
   at OA scale
   

   View Slide

10. Static sequential
    microsimulation
    Source: Lomax, N. and Smith, A P. (2017) Microsimulation for demography. Australian Population Studies, 1(1): 73-85
    

    View Slide

11. Dynamic microsimulation
    

    View Slide

12. Household Change
    Exeter
    Newcastle
    

    View Slide

13. Exeter
    Newcastle
    Population Change
    

    View Slide

14. Source: Pregnolato et al. (2018) A building stock and household
    composition model for the UK. GISRUK, Leicester, 17-20 April.
    Allocation to buildings
    • MasterMap and
    AddressPoint data are the
    input for a spatial
    allocation algorithm that
    classify the building stock
    in terms of residential
    type.
    • Spatial assignment by
    fitting buildings to
    households on the basis of
    the correlation of
    household size and
    number of rooms with
    building area footprint.
    • This allows all buildings
    footprints to be assigned a
    plausible set of household
    characteristics.
    

    View Slide

15. Input to demand models
    Source: Results from MISTRAL energy team, Oxford
    

    View Slide

16. Conclusion (1 of 2)
    • High resolution demographic projections are needed
    for effective modelling of demand across a range of
    sectors
    • Most projection models operate at aggregate scales
    • Microsimulation offers a useful tool to undertake
    modelling at fine spatial scale
    • And the potential to incorporate a range of variables
    

    View Slide

17. Conclusion (2 of 2)
    • Of the two microsimulation approaches:
    • Static sequential is easier to implement but misses
    much of the dynamic processes which go on as
    population changes
    • The fully dynamic model is potentially the best
    performing but very hard to implement. This is
    currently work in progress.
    

    View Slide

18. Nik Lomax
    Andrew P Smith
    AAG New Orleans
    Innovations in Urban Analytics, 12 April 2018
    High resolution demographic
    projections for infrastructure
    planning
    

    View Slide