MultiAgent Simulator for Urban Segregation MASUS A Tool to Explore Alternatives for Promoting Inclus

1
Multi-Agent Simulator for Urban Segregation
(MASUS)A Tool to Explore Alternatives for
Promoting Inclusive Cities

• Flávia F. Feitosa, Quang Bao Le, Paul L.G. Vlek
• Center for Development Research (ZEF)
• University of Bonn
• 3rd ICA Workshop on Geospatial Analysis and
  Modeling, University of Gävle, August 6-7, 2009

2
A Global Urban Age

• Since 2008, the majority of the worlds
  population lives in urban areas

Source UN-Habitat, 2007
3
A Global Urban Age

• Since 2008, the majority of the worlds
  population lives in urban areas

• Cities are not the problem they are the
  solution
  
  J.Lerner
• Need to fulfill their potential as engines of
  development

• Inclusive cities
• Promote growth with equity
• A place where everyone can benefit from the
  opportunities cities offer

4
Urban segregation
A barrier to the formation of inclusive cities
5
Impacts of Segregation

• Obstacles that
• contribute to the
• reproduction
• of poverty

6
Causes of Segregation

• Personal preferences
• Labor market
• Land and real estate markets
• State policies and investments

But How to understand the influence of these
mechanisms on segregation dynamics?
7
The Complex Nature of Segregation

• Segregation displays many of the hallmarks of
  complexity

8
MASUS

• Multi-Agent Simulator for Urban Segregation

• Purpose
• Provide a scientific tool for exploring the
  impact of different mechanisms on segregation
  dynamics
• Virtual Laboratory

9
MASUS Conceptual Model
10
URBAN-POPULATION Module

• Micro-Level
• Household Agent
• Agent profile
• Age, income, education, size,
• tenure status, presence of kids, location
• (b) Household Transition Sub-Model (H-TRANSITION)
• (c) Decision-Making Sub-Model (DECISION)
• Bounded-rational approach ? nested logit
  functions

11
URBAN-POPULATION Module

• Macro-Level Population
• Socio-Demographic State
• Size, income inequality level, and
• other socio-demographic statistics
  (non-spatial)
• (b) Population Transition Sub-Model
  (P-TRANSITION)
• (c) Segregation State
• Product of the spatial location of all
  households
• Depicted by spatial measures of segregation

12
URBAN-LANDSCAPE Module

• Landscape Patch
• Minimal portion of the environment
• 100X100m
• Landscape Patch State
• Land use, infrastructure, land value, number of
  dwellings, distance to roads, distance from CBD,
  slope, type of settlement, zoning variables.
• (b) Urban Sprawl Sub-Model (U-SPRAWL)
• (c) Dwelling Offers Sub-Model (D-OFFER)
• (d) Land Value Sub-Model (L-VALUE)
• (e) Infrastructure Sub-Model (INFRA)

13
EXPERIMENTAL-FACTOR Module

• Specification templates to
• test theories and policies
• Change global variables that affect the
  socio-demographic composition of the population
• Change parameters that drive behavior of agents
• Change structure of DECISION sub-model
• Change the state of urban landscape

14
Process Scheduling
15
Decision-Making Sub-Model
16
Decision-Making Sub-Model

• Nesting Structure of the Model

17
Decision-Making Sub-Model
18
Process Scheduling
19
Urban Population Sub-Models

• Household Transition Sub-Model (H-TRANSITION)
• Rule-based functions representing some natural
  dynamics of the agent profile (e.g., aging)
• Population Transition Sub-Model (P-TRANSITION)
• Keeps the socio-demographic state of the
  population according to levels provided by the
  modeler.

20
Process Scheduling
21
Urban Landscape Sub-Models

• Urban Sprawl Sub-Model (U-SPRAWL)
• Transition phase how many patches become urban?
• Markov chain global transition probabilities
• Allocation phase where?
• Binary logistic regression probability of a
  non-urban patch becoming urban
• Variables urban patches and population density
  in the neighborhood (radius 700m), dist CBD, dist
  roads, slope, zoning

22
Urban Landscape Sub-Models

• Dwelling Offers Sub-Model (D-OFFER)
• Transition phase updates the total number of
  dwellings
• Occupied dwellings (pop) housing stock
• Allocation phase where?
• Linear regression model 1 estimates the patches
  loss of dwellings (expansion of non-residential
  use)
• Linear regression model 2 estimates the patches
  gain of dwellings (new developments)

23
Urban Landscape Sub-Models

• Land value sub-model (L-VALUE)
• Hedonic Price Model Linear regression functions
  to estimate patches land value

• Infrastructure sub-model (INFRA)
• Linear regression model to estimate patches
  infrastructure quality

24
Operational MASUS Model
São José dos Campos, Brazil
City of São José dos Campos
Study Area
São Paulo State
25
Operational MASUS Model
26
Simulation Experiments

• Comparing simulation outputs with empirical data
• Testing theoretical issues on segregation
• Testing an anti-segregation policy

27
Experiment (1) Validation

• Is the simulation model an accurate
  representation
  of the target-system?
• Initial condition - São José dos Campos in 1991
• Import GIS Layers
• Households (Agents) Census 1991, microdata
• Environment (Landscape patches)
• Urban Use, Zoning, Infrastructure, Distance
  CBD, Distance Roads, Land Value, Dwelling Offers,
  Neighborhood Type, Slope.
• Set Variables and Parameters

28
Experiment (1) Validation

• Is the simulation model an accurate
  representation
  of the target-system?
• Run 9 annual cycles
• Compare simulated results with real data (year
  2000)

29
Experiment (1) Validation

• Dissimilarity Index (bw 700m)

Real Data (2000)
Initial State (1991)
Simulated Data (1991-2000)
0.54
0.51
0.31
0.30
0.15
0.19
30
Experiment (1) Validation

• Isolation Poor Households (bw 700m)

Real Data (2000)
Initial State (1991)
Simulated Data (1991-2000)
0.54
31
Experiment (1) Validation

• Isolation Affluent Households (bw 700m)

Real Data (2000)
Initial State (1991)
Simulated Data (1991-2000)
0.15
32
Experiment (2) Inequality

• How does inequality affect segregation?
• Relation between both phenomena has caused
  controversy in scientific debates
• Experiment
• Compare 3 scenarios
• Scenario 1 Previous run
• Scenario 2 Decreasing inequality
• Scenario 3 Increasing inequality

33
Experiment (2) Inequality
Proportion Poor HH
Inequality (Gini)
Proportion Affluent HH
Isolation Affluent HH
Isolation Poor HH
Dissimilarity
Scenario 1 (Original)
Scenario 2 (Low-Ineq.)
Scenario 3 (High-Ineq.)
34
Experiment (3) Poverty Dispersion

• What is the impact of a social-mix policy based
• on the distribution of housing vouchers?
• Experiment
• Compare 3 scenarios
• Scenario 1
• No voucher (baseline)
• Scenario 2
• 200 - 1700 vouchers
• Scenario 3
• 400 - 4200 vouchers

Scenarios
35
Experiment (3) Poverty Dispersion
Dissimilarity
Isolation Poor HH
Isolation Affluent HH

• Scenario 1
• No voucher (baseline)
• Scenario 2
• 200 - 1700 vouchers
• Scenario 3
• 400 - 4200 vouchers

36
Concluding Remarks

• MASUS A Multi-Agent Simulator for Urban
  Segregation
• Explore the impact of different causal mechanisms
  on the emergence of segregation patterns
• Virtual laboratory that contributes to scientific
  and policy debates on segregation
• Three different types of experiment
• Validation comparison with real data
• Theoretical question inequality vs. segregation
• Policy approach poverty dispersion

37
Concluding Remarks

• Suggestions for additional experiments
• Dispersion of wealthy families
• Regularization of clandestine settlements
• Promotion of equal access to infrastructure
• Improve MASUS usability and effectiveness
• Participatory modeling approach
• Feedbacks from potential users

38
Multi-Agent Simulator for Urban Segregation
(MASUS)A Tool to Explore Alternatives for
Promoting Inclusive Cities

• Flávia F. Feitosa, Quang Bao Le, Paul L.G. Vlek
• Center for Development Research (ZEF)
• University of Bonn
• 3rd ICA Workshop on Geospatial Analysis and
  Modeling, University of Gävle, August 6-7, 2009