Multiscale Models of Exurban Land Conversion in Ohio

1
Multiscale Models of Exurban Land Conversion in
Ohio

• Darla K. Munroe
• Hyowon Ban
• Department of Geography
• Center for Urban and Regional Analysis
• Ohio State University

2
Outline

• Definitions and importance of exurbia
• Conceptualizing exurbia as a regional phenomenon
• Modeling development risk
• Multilevel extensions
• Results
• Future directions

3
Definition

• Exurban area
• Urban fringe Built-up area just outside the
  corporate limits of the city (Smith 1937)
• Spectorsky 1955 The Exurbanites
• May contain recreational farms, nursery, mixed
  distribution of old housings and new housing
  developments
• Exurban density 40-325 persons/sq mile

urban
suburban
rural
Simplified conceptual spatial discrimination of
exurban area
4
Websters

• Exurbs are a region, generally semi-rural,
  beyond the suburbs of a city, inhabited largely
  by persons in the upper income groupa person
  living in an exurb especially one commuting to
  the city as a business or professional person.

5
Exurbia vs. sprawl

• Sprawl
• Imprecise - can encompass different urban forms
• Implies attachment
• Exurbia
• Focus on urban-rural interface
• Low density, fragmented, urban-dependent
  development

6
(No Transcript)
7
Motivation

• Exurbanization in Ohio has dramatically increased
  over the last 15 years
• Urban land area has grown twice as fast as
  population
• Exurbanization has detrimental environmental,
  social and economic effects
• E.g., cost of extending services, eroding tax
  base
• Exurbanization is a regional phenomenon
• E.g., landowners move into rural areas to find
  less restrictive policies, lower taxes
• One key driver of growth is the labor market

8
Statement of problem

• In order to understand the relationship between
  the regional economy and exurbanization, we must
  understand how higher-level growth pressures are
  distributed across space

9
Conceptualizing growth pressure

• Movement to high-tech economy
• Flight from blight
• Inconsistent policies
• Dynamics of agricultural sector
• Decreased commuting costs, infrastructure
• Increased wealth
• Cost of living
• Spatial externalities
• Supply-side effects developers

(Sources Boarnet 1996, Brueckner 1998, Byun and
Esparza 2005, Carruthers and Ulfarsson 2002,
Ewing 1994, Irwin and Bockstael 2002, Mieszkowski
and Mills 1993, and Tiebout 1956)
10
Unresolved issues from prior work

• Urban economists cannot explain density
  gradients, changes in commuting time
• Irwin, Anas, Brueckner and others importance of
  agent interactions
• BUT cannot explain certain outcomes
• Speculative real estate bubbles
• Growth despite taxes

11
Potential landowners

• Exurbanites who buy land might consider
• Exurbia as lifestyle choice
• Desire for second/vacation homes
• Home(s) as investment opportunity
• Rental properties
• New migrants to the region

12
Regional perspectives
13
Regions, land use and development

• Carlino-Mills models of population-employment
  flows (Boarnet 1994 Henry et al. 1997)
• Land and labor markets (Johnes and Hyclak 1999,
  Riddel 2001)
• Regional adjustment models and land use
  (Carruthers and Vias 2005)
• Largely aggregated, pattern-oriented

14
The layers of exurbia

• Factors operate on a variety of scales
• Can further delineate into urban- and rural-based
• The pattern of exurbia reflects how urban-based
  and rural-based economies come together (Anas
  2004)

15
Munroe and Irwin
GLOBAL
REGIONAL
LOCAL
EXURBAN
URBAN
RURAL
16
Modeling exurban conversion
17
Irwins (2002, 2004) model

• Focus on timing of development
• Conceptually
• Agricultural landowner has two choices
• Retain land in farming
• Sell to developers
• Land is sold at the point in time when gains from
  conversion (sale of parcel) exceed likely future
  returns from farming

18
Example The point in time when land is sold

• t the point of time when maxA-B

gains
A Discounted net returns from agricultural land
use development in period t
B Development pressure increases and then
decreases
(Compact development may reduce development
pressure on open space and farmland at the urban
fringe )
time
t
19
Analytically

• V sale of parcel
• A returns from ag
• d discount rate
• r interest rate

Irwin and Bockstael, 2002
20
Assumptions of the Irwin model

• Gross returns to development are increasing over
  time
• Population and/or income are rising, and land is
  either fixed, or developable land increasing
  less quickly
• Returns to agriculture remain constant over time
• Land value increases at decreasing rate OR
  development costs are increasing

21
Irwin implementation

• Cox-proportional hazards approach
• To model the variations in timing of residential
  development as a function of observable
  characteristics
• Many of the factors underlying development
  pressure are unobservable, but related to time
  horizon
• Survival model agricultural parcels die when
  optimal time of conversion is reached

22
Our contribution
23
The problem of heterogeneity

• I.e., development pressure varies across space
• Often ignored
• Methodologically
• Estimated coefficients specific to context data
• From a policy perspective
• Perverse outcomes hard to explain increased
  growth along with tax increases

24
Questions about regional / local

• How is the total variation in exurban development
  distributed within a particular MSA?
• How much of the overall variation in exurban
  development is attributable to local
  characteristics?
• How much is attributable to regional factors?
• Does the relationship between regional economic
  growth and exurbanization differ systematically
  across individual counties?
• Can be answered by the multilevel approach

25
Multilevel extension

• Regional level Local level
• Incorporate labor market changes
• Timing of land development varies across MSA
• In part dependent on employment wage
  opportunities
• Advantages
• Regional context labor market changes
  differentially within Ohio
• Local context matters, too suitability of each
  parcel
• Bayesian approaches can extend Cox-proportional
  hazards approach to multilevel

26
Bayesian hierarchical approach

• Advantages
• Specification of varying baseline hazards
• Time-dependent covariate effects
• Random effects across levels of analysis
• Simulation techniques by Markov Chain Monte Carlo
  methods (MCMC)
• Actually not at odds with classical mixed logit
  (Greene 2005)
• Cross-sectional variation in development pressure
• Separate from effect of other local covariates,
  i.e., zoning, school districts, tax regimes
• Local parameters have hyperparameters

27
Empirical application

• Study areas
• Delaware County, OH in Columbus MSA
• Medina County, OH in Cleveland-Elyria-Mentor MSA
• Warren County, OH in Cincinnati-Middletown MSA
• Data
• Parcel-level residential land use data
• Time frame 1988 2003

(Color source Colorbrewer.com)
28
Trends in development
M
D
W
Warren County
Delaware County
Medina County
29
Trends in the three counties
30
Relevant covariates

• Lot size, drainage, floodway, slope
• Distance to activity centers
• Major/minor cities open space
• Accessibility to road infrastructure
• Major roads, interstates
• Policy variation school, tax districts

31
Capturing local economic structure

• Dynamic shift-share analysis
• Track regional (MSA) growth from 1988 2003
• Identifying components
• Industrial Mix the regions relative mix of
  employment across industries
• Regional Shift economic growth in
  region-specific industries (locally important
  industries)

32
Industry Mix
33
Regional Share
34
Trends

• Decline in manufacturing, increase in services
• FIRE important in Columbus
• More dynamic changes in Columbus
• RS in Cincinnati consistently higher than
  Cleveland

35
Time effects

• Huge increase in development pressure starting in
  late 90s
• Low interest rates (not spatially varying)
• Recession beginning 2001
• Lags between business cycle / development

36
Results

• Variations in hazard of development across time
  and space

37
(No Transcript)
38
Drivers of development

• Hazard of development increasing wrt
• Parcel size
• Slope
• Accessibility to Cincinnati, Cleveland
• Distance to interstates

• Hazard of development decreasing wrt
• Distance to minor city
• Distance to major city
• Distance to industrial parcels
• Distance to major roads
• Distance to open space

39
(No Transcript)
40
Next steps
41
Next steps

• Estimating multilevel discrete time to event
  model
• Mandala (2005) fertility in Malawi,
  incorporating community factors
• Start with a prior that covariate effects are
  separable and development pressure constant
• Surface of conversion risk

42
Temporal effects

• Model associations between regional variables and
  local speed of development
• Temporal association could be difficult to
  identify
• Largely an empirical issue

43
What about developers?

• Fortune 500
• Homebuilders 13th most profitable industry
  profit is 9.9 of 2005 revenues
• 2 of the top 5 homebuilders build in Ohio
• 1st Pulte
• 4th Centex

44
Ontology of Irwin model

• Latent, reduced form model of land conversion
• Assumptions
• Development occurs at optimal timing
• Market of myriad buyers/sellers of land
• Can refine those assumptions
• Learning strategies of developers
• Agent heterogeneity

45
Conclusions

• Exurbia is a complex phenomenon
• Exclusively local or regional perspective will
  miss aspects of the process
• What might be sufficient indices of latent growth
  pressure variable?
• Agent heterogeneity interactions among
  landowners, local govt and developers

46
Acknowledgements

• Support for this research was provided by the
  Center for Urban and Regional Analysis (CURA),
  Ohio State University
• Thanks to Paul Hoeffler, Jill Clark, Jim Biles,
  and Elena Irwin