1. PROJECT
DEFENSE
THE IMPACT OF A LIGHT RAIL
SYSTEM (EXISTING BLUE LINE) ON
SINGLE FAMILY PROPERTY VALUES
IN MECKLENBURG COUNTY, NC, FROM
1997 TO 2008
SisiYan 2009
Master of Arts in Geography
University of North Carolina at Charlotte
Department of Geography and Earth Science
Committee Members: Dr. Eric Delmelle, Dr. Mike Duncan and Dr. Harrison Campbell
4. 1984---the Charlotte-
Mecklenburg Planning
commission made its
first recommendation;
1998, Dec---tax
voted, the planning for
the South Corridor to
Pineville commenced;
2005, Feb26---
groundbreaking;
2007, November 24---
Opened
5. Research Questions:
how much is the property value
change as it proximity to rail
transit in Charlotte area?
Was there such impact during
plan time? How have this
relationship changed over time?
6. Content Table
•Introduction
•Literature Review
•Research Design & Location Theory and
Hypotheses Transit Capitalization
•Study Area & Data
Hedonic Price Studies
•Method
•Results & Discussion Empirical Studies on
•Conclusion & Future Transit Impact
Study
7. Literature Review
Location Theory
Von Thünen 1826 (land use theory)
Different land use will be adopted accordingly in order
to maximize the overall profits. land that is closer to
the market place will bear less transportation costs
and therefore has higher value.
Alonso 1964, Muth 1969:(bid for rent)
higher land value appears in a shorter distance to
center and this rent gradient will decline nonlinearly
as distance to center increases.
“ good accessibility results in higher property values “
8. Hedonic Price Model
Knaap (1998) summarized:
Property character: the size, age and quality of any
structure, etc;
the location character: distance to CBD, transit and
other amenities.
the neighborhood character: median household
income and crime rate, etc.
Sale _price =f (Pr, H, L, N)
9. Empirical Studies
Light rail in Portland, Oregon (Lewis-Workman
and Brod, 1997)
on average, property values increase by $75 for every
100 feet closer to the station
Metrorail in Miami, FL (Gatzlaff and Smith, 1993)
weak evidence that there was any major effect on
residential values because of the rail
Rapid Transit in Chicago (McMillen and
McDonald, 2004)
the housing market anticipated the opening of the line and
house prices have been affected by proximity to the
stations six years before its construction
11. Study Time Frame
T1 T2 T3- T4
• Pre- • Planning • Rail • Rail
Planning period; Construction Operation
period; From 1999 period; period;
From 1997 to 2004 • From 2005 to After
to 1998 2007 Nov.1st
2007 till
July 2008
14. Charlotte
Since the 1980s, Charlotte has been one of
the nation’s fastest growing urban areas.
Between 1980 and 2005, Charlotte grew from
the 47th to the 20th most populated city in
the United States (Charlotte Chamber).
Due to the development of the banking
industry, Charlotte became a financial city
attracting many new businesses
16. Data Sources
Mecklenburg County & UNC Charlotte Urban
Institute
Charlotte Area Transit (CATS)
Federal Housing Finance Agency
US census
other secondary data generated by Geographical
information technology (GIS).
17. Table 3 Descriptive Statistics
Minimum Maximum Mean Std. Deviation
sales_pric 10000.00 992000.00 197997.82 146058.84
age 1.00 108.00 46.49 21.31
heatedarea 480.00 7003.00 1722.25 743.16
height 1.00 3.00 1.32 0.48
NUM_fire 0.00 4.00 0.67 0.49
qality_building 0.00 4.00 1.47 0.89
fullbaths 1.00 6.00 1.69 0.71
bedrooms 1.00 9.00 3.12 0.64
units 1.00 2.00 1.00 0.05
lnheatarea 6.17 8.85 7.38 0.38
lnnetdis 6.27 9.93 8.42 0.53
t1lnnetdis 0.00 9.93 2.11 3.67
t2lnnetdis 0.00 9.92 3.39 4.14
t3lnnetdis 0.00 9.93 1.72 3.40
t4lnnetdis 0.00 9.93 1.20 2.95
Valid N 6381
Note: t(i) Lnnetdisrepresents the ln_net_dis (in feet) at t(i) (i=1,2,3,4) time period
20. Methods:
hedonic regression model for four time periods:
Model 1:
Sale _price =f (Pr, H, N(i))
Model 2:
Sale _price =f (Pr, H, BG(i))
Specify model:
Ln(ad_sale_ price) =β0+βi * hi +βj * ln_net_distance+βk * Dumk +
εi
Where, dependent variable is the natural logarithm of the adjusted
sales price; hiis a vector of asset-specific characteristics of the
properties; ln_net_distance is the logarithm of proximity variable;
Dumkis spatial dummy variables; βistands for the coefficients of each
independent variable;
21. Spatial Dependence
Neighborhood
Boundary
Moran’ s I
0.85
Block group
Boundary 0.8
0.75
Moran's I t1
0.7
t2
t3
0.65
t4
0.6
0.55
300 500 600 650 700 800 900 1100 2000
Threshold Distance (feet)
22. Variables Discussion
Sales value vs. assessed value
Network distance vs. Straight-line Distance;
Variable List
30. Conclusions
Contradictory to many studies, single family housing
value in Charlotte area tend to increase value as distance
to rail increases
Comparing across four time periods, pre-
planning, planning, construction and operation, rail
operation diminish the proximity disadvantage that
appears at the station area
31.
32. Future Studies
Apply model to other available property types such
as multiple family and commercial
Analyze the impact of rail when the line is completed
Integrate spatially-explicit regression models such as
geographical weighted regression
Local patterns in residuals
Divide study time period according to station plan
time
33. Acknowledgements
Thanks for Eric’s advice from Idaho to Charlotte
Thanks for Mike’s great help and guidance
through this study
Thanks for Harry’s support
Thanks for Tom Ludden’s data support
Thanks for Paul McDaniel's great tolerance
during editing my ‘professional’ Chine-
glishwriting
Thanks for Amos’s Coding support
Thanks you all for coming today
35. References Selected:
Al-Mosaind, M.A., Dueker, K.J., Strathman, J.G.
(1993), "Light rail transit stations and property values: a
hedonic price approach", Transportation Research
Record, No.1400, pp.90-4.
Alonso, W. (1964). Location and land use: Toward a
general theory of land rent. Cambridge, MA: Harvard
University Press.
Bajic V (1983). The effects of a new subway line on
housing prices in metropolitan Toronto. Urban Studies
20: 147–158.
Duncan, Michael (2007) The Conditional Nature of Rail
Transit Capitalization in San Diego, California.
Dissertation No. D07-003
36. Variables Description Data Sources Justification
PROPERTY VALUE (dependent variable)
Amount($) for which the single family
the sales price generally reveals the
property was sold during the study time the Property Ownership Land Records
value of the property. (Bowes and
Ln_ad_Price period. Dollar values are adjusted to the third Information System (POLARIS)
Ihlanfeldt, 2001; Voith,1993;Al-
quarter of 2005 based on HPI(Housing Price Federal Housing Finance Agency
Mosaind et al,1993)
Index).
RAIL PROXIMITY
semi-log of network distance(in feet) to the real access distance.(Duncan, 2007;
Ln_Netdis Calculated using GIS
nearest rail station Landis et al.1995)
PROPERTY CHARACTERISTICS
age of the structure(in year) 2008 age may affect the price of the
Age POLAIRS
substract building year building.
squared age may capture the
Age2 squared age POLAIRS nonlinear relationship between
age and price (Coulson, 2008)
semi-log of heated area(in square feet) of
ln_HeatedArea POLAIRS same as above
the property
Fullbaths number of bathroom in the unit POLAIRS same as above
Bedroom number of bedroom in the unit POLAIRS same as above
Actype (Ac01, Ac02, Primary type of air conditioning system
POLAIRS same as above
Ac03, Ac04,) used (4 categories of AC)
the quality of the structure(below average
Qality_bui POLAIRS same as above
to excellent, 1-5)
UNITS Number of living units in the structure POLAIRS same as above
HEATEDFUEL (Fuel01, Primary type of fuel used for heating (5
POLAIRS same as above
02, 03, 04, 05, ) categories of Fueltypes)
HEIGH story height POLAIRS same as above
NUM_FIRE number of fireplace POLARIS same as above
37. LOCATIONAL & NEIGHBORHOOD CHARACTERISTICS (based on two scales)
Consider the
whether or not the property is neighborhood boundary
City of Charlotte
F(i) within a neighborhood as dummy variables to
Quality of life study and GIS
i(0,1,6,900,etc) control for loccation and
neighborhood characters
Consider the block
whether or not the property is group boundary as
Dum(i) within a block group i(0- US Census and GIS dummy variables to
34,etc) control for location and
neighborhood characters
38. Table 4 Price Statistics for four time periods
Note: ad_price is the adjusted price that is calculated by House Price Index.
Time_Preiod avg_ad_price min_ad_price max_ad_price N
t1 197,950 13,422 1,133,820 1,592
t2 206,720 10,527 1,007,040 2,568
t3 213,300 15,000 990,000 1,308
t4 227,840 13,849 845,585 913