1. Uncertainty in socioeconomic forecasts
Todd Graham
Metropolitan Council Research

2. Why forecast? Provides a reasonable basis for planning
local comprehensive planning
regional system planning
Engages stakeholders in addressing growth issues
Helps us understand trends and forces
Forces us to articulate our expectations
11/10/2018

3. Forecast certainty is not possible
DF = Development Framework
SD = State Demographer
11/10/2018

4. Many futures are possible
Many scenarios are possible
What do we imagine is the end-state?
What path takes us there?
Starting assumptions that will constrain the range of possibilities
Narrowing from the possible to the probable
11/10/2018

5. Where does forecasting come in?
Forecast modeling is a system analysis
To represent a set of variables over time
And to represent the dynamics and relationships that move those variables
Probable range of futures
Or the most probable future…
Given a basket of system dynamics, trends, policies, other assumptions
11/10/2018

6. Are multiple forecasts possible?
Probable range of futures
Or the most probable future…
Given a basket of system dynamics, trends, policies, other assumptions
11/10/2018

7. Twin Cities Population Possibilities Range
Hi Life Exp.
Lo Life Exp.
Lo Migr.
All High
Hi Fert.
Hi Migr.
All Mid
Lo Fert.
All Low
Thousands
11/10/2018

8. The most probable future(s)?
System dynamics and trends
Can be tweaked as appropriate by forecaster
Or trends can be endogenously modeled, or loaded in from other related models
Policies are variable
Different scenarios to explore policy options
Policymakers decide; forecasters assist
Result is a policy-based forecast – the desired future
11/10/2018

9. Challenges and opportunities
Improvement of modeling practices
Integration or coordination of parallel forecast efforts
Engagement of policymakers, planners and publics
11/10/2018

10. The Future of Forecasts at Met Council
Todd Graham
Metropolitan Council Research

11. ??? Metropolitan Council’s current model REGIONAL Jobs Households
Population
LOCAL
Land use, current and planned
Current model does not consider spatial interactions
Currently, no feedback between land use and transportation dynamics
???
accessibility
trip generation
Transportation System
Demand distribution  Mode choice 
Network assignment
11/10/2018

12. Complex Metro & Urban Dynamics: Elements and Interactions
REGIONAL
Economy and labor market dynamics
___________
LOCAL
REGIONAL
Jobs
Population
Households
production & consumption
LOCAL
development & occupancy
Land and
floorspace
price
signals
Spatial
interaction
Social & environmental outcomes
accessibility
trip generation
Transportation System
Demand distribution  Mode choice 
Network assignment
Acknowledgment:
Modified from JD Hunt, et al. (2005)
Acknowledgment:
Modified from JD Hunt, et al. (2005)
11/10/2018

13. Expected forecast models workflow
A regional economic model for economic activity, employment, and population
Preferred model: Regional Dynamics (ReDyn.com)
A demographic model for parsing population into households
Preferred model: ProFamy (ProFamy.com)
A land use model for allocating future land use, households and employment to the local level
Preferred model: Citilabs Cube Land
Travel demand model
Currently in use: Citilabs Cube Voyager
11/10/2018

14. Program Objectives Land economics and geographic science validity
Platform for the prediction of likely distributions of development and activity – given a set of rules, or given a set of represented behaviors or dynamics
Coordination/integration with Travel Demand Modeling (TDM) and ES capital planning
Model land use dynamics and transport network together – to better represent trends
11/10/2018

15. Goals developed via Needs Assessment Workshops
A model that balances the need for transparency with the need for realism
Able to test a range of policy scenarios
A model that provides information on the interaction of the physical environment and development dynamics interact
Geographic scope and level of detail necessary for regional systems planning
Flexibility to forecast short-term, long-term, and “build-out”
11/10/2018

16. Cube Land
2010
11/10/2018

17. No – modeled separately
Market-based integrated models evaluated against Met Council Needs Assessment
Theoretical foundations:
Understandable methodology, with explanation
Mostly
Traceability of results and ability to perform sensitivity tests
Yes
Basis in valid regional and urban development theory
Capacity to model and test a wide range of policies
Demographic capabilities of model
No – modeled separately
Spatial interaction of physical environment and development:
Flexibility to incorporate variety of layers into model
Allocate growth based on transportation network and accessibility measures
Socioec-land use model outputs used as travel demand model inputs
Temporal resolution: Ability to forecast 30 years, at 5-10-year intervals
Forecast for very-long-term: 50 years
TBD
Geographic granularity: Micro-level simulation results (parcel level)
Varies by model
11/10/2018

18. Evaluated against Hunt, Kriger, Miller (2005) review of best practices
Theoretical foundations:
Real estate market modeled with endogenous pricing – i.e. demand, supply, prices are interdependent and can adjust
Yes
Model accounts for key subsystems of region – networks, land use, built environment, activities, travel
Capacity to model and test a wide range of policies: Provides measures of benefits and costs of policy alternatives
Spatial interaction of physical environment and development:
Framework for modeling interaction between land use and transportation
Transit representation and sensitivity
TBD
Outputs include predicted land use by type, built environment, segment detail on households and employment industry sectors
Geographic granularity: Analytical units at finest-possible level of detail -so as to maximize behavioral simulation
Varies by model
Feasibility: Parsimonious data requirements
CubeLand – Yes
Manageable implementation requirements (given timeline and budget)
11/10/2018

19. Cube Land – a market based model
Equilibrium represented by simultaneous solution of three inter-dependent problems:
Location of real estate consumers
Supply of real estate
Rents and values at market-clearing equilibrium
11/10/2018

20. Background on Martinez’s Modelo de Uso de Suelo de Santiago
Martinez, Franisco; and Pedro Donoso. “MUSSA 2: A Land Use Equilibrium Model Based on Constrained Idiosyncratic Behavior of Agents in an Auction Market.” Paper at TRB Annual Meeting, January pages.
“MUSSA – Land Use Equilibrium Model.” February 2009 presentation at presentationsSeminaires/MUSSA_Martinez09.pdf
“MUSSA – Its Basis.” 4 pages. Website at
11/10/2018

21. Cube Land – a market based model
On demand side, households (h) buy or rent real estate type (v) at certain locations (i)
Neighborhood choice (location i) determined by income and willingness to pay:
Bhvi = Ih – {f(Uh–zvi)}
Where Uh is typical housing utility for an “h” household
Where zvi represents package of amenities, neighborhood characteristics
Better package  greater willingness to pay
Max (Bhvi – rvi)
Subject to available budget of “h” household
11/10/2018

22. Cube Land – a market based model
On supply side, developers (j) will offer housing & built space in quantities (S) of certain type (v) at certain locations (i) in order to maximize profit
Max {SviJ* (rvi – cviJ)}
Subject to regulations at location “i”
And all households in region are matched with housing
Predicted location choices and predicted supply are calculated with MNL equations (i.e. choice probabilities)
11/10/2018

23. 11/10/2018

24. Integrated modeling
Travel times, accessibility and networks are updated and inform socioeconomic/land modeling at each 5-year step
Base Transport
Model
Base SE-LU
Updated Network
& Access
SE-LU 2010
SE-LU 2015
Updated Network
& Access
SE-LU 20##
Updated Transport
Model
11/10/2018

25. Policy and regulation constraints
Permissible land uses
Housing unit density min/max
Building height max or FAR max
Protected land and planned parks/reserves
GIS coverage of aquifer depletion
Wastewater system capacity constraints?
11/10/2018

26. Cube Land – a market based model
Equilibrium represented by simultaneous solution of three inter-dependent problems:
Location of real estate consumers
Supply of real estate
Rents and values at market-clearing equilibrium
11/10/2018

27. Cube Land – a market based model
Cube Land outputs not only what land will be developed – but also what types of housing – and prices for real estate zones
11/10/2018

28. Integrated modeling preferred
Are these applications possible?… With these tools…
Travel Model Alone?
Integrated Modeling Program?
Land use/transportation interactions: Assess interactive effects of transportation system on land uses, and vice-versa – either “constrained” by land use plans or “free-market” No
Yes
Land Use Analysis: Predict amount and locations of land uses (residential, commercial, industrial, and employment)
Smart Growth: Analyze the effects and benefits of Smart Growth strategies (infill and TOD in coordination with transit service)
Poorly
Jobs/Housing Balance: Based on incomes of residents and employees in relation to housing prices
Planning strategies: Assess traffic-related effects/benefits of urban growth boundaries, growth management strategies, impact fees
Transportation System Management: Effects of land uses and help set priorities among competing projects. A consistent approach for comparing potential improvements or alternatives.
Source: Johnston, R; and M McCoy. (2006): Assessment of Integrated Transportation-Land Use Models: Final Report. Online at
11/10/2018

29. Complex Metro & Urban Dynamics: Elements and Interactions
REGIONAL
Economy and labor market dynamics
___________
LOCAL
REGIONAL
Jobs
Population
Households
production & consumption
LOCAL
development & occupancy
Land and
floorspace
price
signals
Spatial
interaction
Social & environmental outcomes
accessibility
trip generation
Transportation System
Demand distribution  Mode choice 
Network assignment
Acknowledgment:
Modified from JD Hunt, et al. (2005)
Acknowledgment:
Modified from JD Hunt, et al. (2005)
11/10/2018

30. Challenges and questions
Are the forecasts responsive to economics, market conditions, and urban dynamics?
Are the forecasts responsive to – or realistic considering – policies and plans?
If so, how?
Are the transportation forecasts responsive to future land use and socioeconomics?
And vice verse?
11/10/2018

31. Integrated modeling
Travel times, accessibility and networks are updated and inform socioeconomic/land modeling at each 5-year step
Base Transport
Model
Base SE-LU
Updated Network
& Access
SE-LU 2010
SE-LU 2015
Updated Network
& Access
SE-LU 20##
Updated Transport
Model
11/10/2018

32. Integrated Models - Paths of Advancement
Travel
Demand
No Transit
Transit
Land
Advanced
Aggregate
Model
Activity-based
No Mode Split
Logit Model Split
Use
Model
Land Capacity,
Trends, Judgment
Met Council in 2008
Non-market-based
land allocation
Land allocation
with price signals
Fully integrated
market-based
model
Met Council in 2010
Ideal Model
Path of advancement
Source: Miller, EJ, et al (1999): Integrated Urban Models for Simulation of Transit and Land Use Policies.

33. Integrated modeling as a policy ideal
Transportation Policy: SAFETEA-LU and ISTEA
Coordination of land use and transportation planning
NEPA and Clean Air Act
Land development patterns must be consistent with regional transportation plan
11/10/2018

34. Uncertainty in socioeconomic forecasts
Todd Graham
Metropolitan Council Research