STEP3: Micro-simulation-Based Small Area Demographic Forecasts for Clark County
Stewart BerrySrinivasan SundaramHoward Slavin
Caliper Corporation2007
Introduction
• This project entailed development
and application of STEP3- a disaggregate household microsimulation model
• STEP3 produces small area population forecasts for Clark County
• STEP3 can be used for planning, impact assessment and monitoring
STEP3 Framework
HOUSEHOLD BEHAVIOR(Simulates behavior for individual households and
persons)
Zone Data(Employment and Landuse
Data, Transportation and Accessibility)
POPULATION SYNTHESIS(Generates Household and Person databases that are
representative of the population)
POPULATION PROGRESSION(Progresses population through vital life events)
- Population Progression- Workforce Participation- Retirement status
Model Component
Model Flow
Input/Output File
Input/Output Flow
LANDUSE MODELING
- Employment Location- Housing Location
Zone Data: Demographic splits by Household
Size, Income, Age of Head of Household, etc.)
PUMS data: Individual Household Person Census Records
Synthetic Person File
Synthetic Household File
Lifestyle and Mobility Decisions
- Residential Location- Workplace Location
Synthetic Household FileSynthetic Person File
Population Progression
Aging, Mortality and Births
• Age by 1 year• Education of children is increased• Income and wages increase• Death rates are applied• Birth rates are applied
Household Formation
• Kids Leave home at age 22
– Vehicles, employment & income are calculated
• Divorces happen– Income & vehicles are split; children are
assigned using custody probability
• Marriages happen– Single men are identified & potential
brides are searched for based on age
Migration
• Regional in- and out- migration is
modeled using rates from IRS tax returns
• Intra-county migration is modeled using rates from the 2000 Census
Labor Force Participation
• Worker status
– Determined by gender, age, race, marital status & children by age
• Retired status– If aged 65+, retirement status is
determined by gender, age & household structure
• Unemployed status– Determined using published Clark County
rates
Land Use Modeling
External Inputs
• The user can add planned residential
and employment buildings:– Construction year– The number of owner/renter units– The number of jobs in 7 sectors:
• Hotel• Office• Industrial• Regional Retail• Community Retail• Neighborhood Retail• Other Non-Retail
Post-2000 Development Layer
Undevelopable Land Constraints
• Undevelopable land restricts growth:
– Military installations– Airports– Water bodies– Parks– Steep gradient – Constrained lands
Residential Cell Growth
• The user can increase or decrease
settlement sprawl and density• A cell can be developed when it:
– Has developable land– Has 2 neighboring cells with 919 people in
each– Is not a group quarters cell
Cell Characteristics Influencing Urban Growth
Employment Seeds
• Non-retail employment grows using:
– Future land-use layer- Growth rates
• Retail employment grows using: – “Hot-spots” that identify areas where
there is high population but little retail
Locational Choices
Hotel/Gaming Industry Workers
• Choose work zone first• Employment preferences:
– CBD – Strip– High employment zones
• Residence preferences:– Income– Owner or renter status– Travel time to work– Number of units available
All Other Workers
• Choose residence zone first• Residence preferences:
– Income– Owner or renter status– Average travel time to work– Number of units available
• Employment preferences:– CBD – Strip– Closeness to home zone– Vehicle & transit travel times & costs
Socio-economic Trends in Clark County
Time-Series of Socio-Economic Data (1990-2005)
• High population growth• Trend of higher percentage of
workers• But population is still growing faster
than the employed• Migration is a major factor in growth
Time Series of Migration
Time-Series of Socio-Economic Data
020000400006000080000
100000120000140000160000
1990
1992
1994
1996
1998
2000
2002
2004
Year
Per
son
s
Inmigration
Outmigration
Previous Population Projections & Estimates
• Five sets of projections and
estimates– Caliper Tract Estimates– IRS tax returns (highest)– State Demographer– REMI– RTC-SNWA-CCCP-CBER (highest)
• Overall trends are consistent• STEP3
– Scenarios model these upper and lower bounds
Population Projections and Estimates
0
500000
1000000
1500000
2000000
2500000
3000000
3500000
4000000
1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 2032 2034
Year
Popu
lation
Caliper Tract Estimates
Comp. Plan
State Demographer
IRS
REMI
Development of the Spatial Database for Analysis
• 1km grid cells• The land use types are:
– Hotel– Office– Industrial– Regional Retail– Community Retail– Other Retail– Other Non-Retail– Renter Occupied– Owner Occupied
Model Structure
Input Data Model Database
Census aggregate andPUMS data: Individual
Household & Person Census Records
Zone Data: Demographic splits by Household Size, Income, Age of Head of Household; Employment and Land Use Data; Employment & Housing Location; etc.
County Business Patterns Employment File
Synthetic Person File
Synthetic Household File
GIS Overlays: Slope/gradient, physical features, human
landmarks
Data Integration in TransCAD
Census Data
Aggregate Zonal Disaggregate Household Disaggregate PersonPopulation Number of people in
householdRelationship to householder
Dwelling Units Place of residence GenderOccupied Housing Units by Tenure
Number of automobiles Race
Median Rent Household income AgeMedian Owner Costs Rent or own EthnicityAggregate Household Income 1999
Group Quarters Type of Education
Occupied Housing Units by Tenure by Car Ownership
Years of schooling
Household Income Employment status
Occupied Housing Units by Tenure by Householder Age
Hours worked last week
Occupied Housing Units by Household Size
Place of work
Group Quarters Industry
Total person's income
Wages or salary
Nonfarm self-employment
Subfamily Relationship
Marital Status
Las Vegas Valley Landmarks
2004 Las Vegas Landuse Types, Gradients & Undevelopable Land
Road Network (2000-2005)
Forecast Results
Results Overview
• Four STEP3 scenarios – High growth with extensive urban dispersion – High growth with constrained urban
dispersion – Lower growth with extensive urban dispersion – Lower growth with constrained dispersion
• The upper bound forecasts exceed the CCCP population forecast
• The lower bound trends are more inline with REMI and the State Demographer
Population Projections
Population Projections
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
3,000,000
3,500,000
4,000,000
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
Year
Po
pu
lati
on
Comp. Plan
State Demographer
REMI
Caliper LBF
Caliper UBF
Upper Bound Forecasts
• Increases in:
– Population– Households– Movers and immigrants– Workers, retirees and non-workers, while
unemployment stabilizes at about 4%
• Constrained scenario growth: – Moapa– Goodsprings (east & south)– Searchlight & CAL-NEV-ARI– Primm-Roach-Borax corridor– Boulder City (south)– Las Vegas (north, south-east, south-west,
west)
Upper Bound Workers
Workers
0
100,000
200,000
300,000
400,000
500,000
600,000
2000
2003
2006
2009
2012
2015
2018
2021
2024
Year
To
tal
Hotel
Office
Industrial
Regional Retail
Community Retail
Neighborhood Retail
Other Non-Retail
New Populated Cells 2000-2025 (upper bound; black are post-2000)
New Populated Cells 2000-2025 (upper bound dispersion; black are post-2000)
Lower Bound Forecasts
• Increases in:
– Population– Households– Movers and immigrants– Workers, retirees and non-workers, while
unemployment stabilizes at about 4%
• Constrained scenario growth: – Moapa– Goodsprings (east)– Searchlight– Primm-Roach-Borax corridor– Boulder City (south)– Las Vegas (north, south-east, south-west,
west)
Lower Bound Workers
Workers
050,000
100,000150,000200,000250,000300,000350,000400,000450,000
2000
2003
2006
2009
2012
2015
2018
2021
2024
Year
To
tal
Hotel
Office
Industrial
Regional Retail
Community Retail
Neighborhood Retail
Other Non-Retail
New Populated Cells 2000-2025 (lower bound; black are post-2000)
New Populated Cells 2000-2025 (lower bound dispersion; black are post-2000)
Model Interface
Description of Model Stages
• Stage 1: Population Synthesis • Stage 2: Zone and Network Processing• Stage 3: Population Progression• Stage 4: Landuse Modeling• Stage 5: Residential Location• Stage 6: Work Location• All Stages: Feedback Loop
Description of Default Scenarios
• Scenario 1: Upper Bound Population
Forecast with Constrained Cell Growth• Scenario 2: Lower Bound Population
Forecast with Constrained Cell Growth• Scenario 3: Upper Bound Population
Forecast with Unconstrained Cell Growth• Scenario 4: Lower Bound Population
Forecast with Unconstrained Cell Growth
Time Series Visualization Tool
• Once STEP3 has been run, the zonal
output can be displayed as a time series:
Future Land-use Database Editing Tool
• The geography of the future land use
layer can be modified:
Land Use Evolution
Case StudyImplications for HAZMAT Exposure
Discussion
• Yucca Mountain Repository for
radioactive waste• Proximity to Clark County of the
facility and transportation routes• STEP3 is used to explore the
projected population that could be exposed
Proposed Nuclear Waste Routes
Projected Population Exposure
• 800 meter (½ mile) buffer• Higher population along the highway
than the rail line• More people exposed in the
dispersion scenarios at the beginning, but are less than the constrained towards 2025
Exposed Population Along Highway Corridor (Clark County)
Exposed Population Along Highway Corridor
0
20000
40000
60000
80000
100000
120000
2000
2003
2006
2009
2012
2015
2018
2021
2024
Year
Po
pu
lati
on LBF Hywy
LBF Disperse Hywy
UBF Hywy
UBF Disperse Hywy
Exposed Population Along Rail Corridor (Clark County)
Exposed Population Along Rail Corridor
0
50000
100000
150000
200000
250000
300000
2000
2003
2006
2009
2012
2015
2018
2021
2024
Year
Po
pu
lati
on LBF Rail
LBF Disperse Rail
UBF Rail
UBF Disperse Rail
Conclusion
• The settled area has grown
enormously and continued sprawl is expected
• Even though predictions are uncertain, planning is aided by forecasts that predict where people live and work
• Household microsimulation, with all its limitations, can be a practical tool for demographic and land use forecasting