Modelling the Impact of Accessibility to Services on House Prices: A Comparative Analysis of Two Methodological Approaches

1
Modelling the Impact of Accessibility to
Services on House PricesA Comparative Analysis
of Two Methodological Approaches

• François Des Rosiers, Marius Thériault Yan
  Kestens
• European Real Estate Society
• 10th Annual Meeting, June 11-13, 2003

2
Introduction

• This paper is an attempt to bridge the gap
  between, on the one hand, the mobility behaviour
  of households and their perception of
  accessibility to urban amenities and, on the
  other hand, house price dynamics as captured
  through hedonic modelling
• It consists of an empirical test of the impact of
  accessibility on house prices, whereby hedonic
  modelling is applied to some 952 single-family
  houses sold in Quebec City between 1993 and 1996
• Two accessibility measures are compared the
  former measure is based on simulated travel times
  to nearest amenities aggregated through factor
  analysis (PCA)
• The latter rests on perceived accessibility
  indices obtained via a fuzzy logic approach
  applied to observed trips patterns derived from
  the 2001 QMA O-D survey

3
Hypothesis and objectives

• Our hypothesis is that different people having a
  heterogeneous perception of space, they will
  adjust their willingness to pay for additional
  centrality/accessibility when choosing their home
  location depending on their needs and preferences
  
• The main objective of this paper is to test
  whether perceptual indices of accessibility are
  actually internalized in housing prices
• Secondary objectives are
• Testing for the marginal contribution to value of
  access to various amenities work places,
  schools, shops, groceries, health care centres,
  restaurants, leisure places
• Testing for the differential impact of
  accessibility among types of individuals and
  households
• Testing for the way life cycle and income impact
  upon the perception of accessibility and is
  translated into house prices

4
Accessibility and House Values (1)

• Traditional urban models are currently based on
  the centrality concept (distance decay function)
  and on accessibility to the CBD (monocentric
  model)
• McMillens (2003 Chicago) decades of urban
  sprawl in North American cities did not weaken
  the prominence of the centrality concept
• Impact of proximity and accessibility to services
  on property values Guntermann and Colwell 1983,
  Colwell, Gujral and Coley 1985, Colwell 1990,
  Grieson and White 1989, Sirpal 1994, So et al.
  1997, Smersh and Smith 2000, Des Rosiers et al.
  (1996, 2001 2003 Quebec City)
• Not all authors though agree on the actual
  influence of accessibility upon house prices and
  residential mobility (McGreal et al. 1999
  Belfast Bordeaux)

5
Accessibility and House Values (2)

• Polycentric cities mere Euclidean distances to
  the CBD falls short of integrating all relevant
  aspects of accessibility (Jackson 1979, Dubin and
  Sung 1987, Niedercorn and Ammari 1987, Hoch and
  Waddell 1993)
• Despite use of minimum travel time and walking
  distance (Bateman et al. 2001), the faulty
  specification of accessibility descriptors may
  explain rather poor performances
• Travel surveys, commuting patterns and
  accessibility to jobs and houses
• Levinson (1996 Washington, DC) suburbanization
  of jobs maintains stability in commuting
  durations despite rising congestion and
  increasing work and non-work trip making and
  length
• Helling (1996 - Atlanta) Effect of residential
  car accessibility to jobs on the quantity and
  nature of travel by men and women - Accessibility
  do not affect everyone while gravity indices only
  provide partial information
• Srour et al. (2002 - Dallas-Fort Worth) Apply
  both general and specific accessibility indices
  to the modelling of residential and commercial
  markets - While common accessibility measures do
  not perform that well, job accessibility indices
  impact positively on residential land values

6
Database Modelling Approach

• Database hedonic modelling applied to 952
  single-family houses sold in Quebec City between
  1993 and 1996 - sale prices range from 50 000 to
  460 000 (Can.)
• High variance on prices use of a multiplicative
  functional form (ln of sale price ln SP)
• Three steps
• Model 1 Ln SP f Property Specifics,
  Inflation, Taxation
• Model 2 Ln SP f S, I, T, PCA of travel times
  to nearest amenities
• Model 3 Ln SP f S, I, T, PAI Perceived
  Accessibility Indices
• Phone survey among buyers revealed that
  accessibility to services, jobs, schools,
  highways and transit networks was an important
  criteria for choosing new neighbourhoods
• Model 3a Ln SP f S, I, T, PAI Buyers Age
• Model 3b Ln SP f S, I, T, PAI Buyers
  Income

7
Factor Analysis - PCA (Nearest Amenities)

• Step 1 compute 15 travel times (car and walking)
  to the nearest local regional amenities
  primary high schools, colleges, universities
  regional, neighbourhood local shopping centres
  CBD
• Step 2 PCA - extract 2 principal components
  using Varimax rotation
• Factor 1 access to nearest regional-level
  services (42 of variance)
• Factor 2 access to nearest local-level services
  (34 of variance)
• Already used by Des Rosiers et al., 2000 Quebec
  City
• Mutually independent factors help control
  multicollinearity
• Step 3 Model 2 - Factor scores are substituted
  for access attributes

8
Factor Analysis - PCA (Nearest amenities)
Local-level services
Regional-level services
9
Modelling Perceptual Accessibility

• Model 3 Accessibility indices were computed for
  significantly different types of persons and
  activities using

where  Ai Raw suitability of residential
location i (sum of suitable opportunities) Sij
Suitability index of travelling from residential
location i to activity location j Total
number of potential activities at location
j where Ai Accessibility index of
residential location i relative to the most
suitable place
10
Perceived Accessibility to Restaurants
C50 5,3 min. C90 12,6 min.
11
Analysis of Results (1)
All models do perform well in spite of remaining
spatial autocorrelation among residuals
12
Analysis of Results (2)
Size and Age coefficients are strengthened. Tax
rate effect declines. This suggests structural
spatial links among these variables and urban
form.
Model 1 All coefficients highly significant and
consistent with expectations. Prominence of age,
size and taxation
13
Analysis of Results (3)
Models 4, 9 and 10 Accessibility to schools and
health care facilities for families as well as to
restaurants exerts strong influence on
prices. Perceived accessibility indices overcome
centrality.
14
Analysis of Results (4)
Model 3a People aged 35-54 are willing to pay a
substantial market premium to locate at a
reasonable travel time from their work
place. Model 3b The higher the household
income, the stronger the propensity to lessen
work-trip duration under an income constraint,
households trade-off longer commuting trips for
cheaper land.
15
Conclusion Research Agenda

• The two sub-hypothesis of this research were
• 1 Various types of persons experience different
  constraints and are not equally willing to travel
  in order to reach various kinds of activities,
  meaning that they have an heterogeneous
  perception of space
• 2 Households will adjust their willingness to
  pay for additional centrality/accessibility when
  choosing their home location depending on their
  needs and preferences
• Both sub-hypotheses are supported by empirical
  results, suggesting the accessibility concept
  might be less straightforward than is usually
  considered in the literature
• The physical, absolute concept of accessibility
  ought to be complemented by behavioural
  approaches integrating people wills and needs in
  their valuation of urban space
• Considering they are a paramount determinant of
  location choices and property values,
  accessibility/centrality issues deserve being
  further investigated using novel tools, including
  travel and activity surveys and modelling