Inner City Business Sector in Phnom Penh, Cambodia:
Exploratory Spatial Data Analysis
Geoffrey Caruso
Geography and Spatial Planning Research Centre, University of Luxembourg
Thomas Kolnberger
Department of History, University of Luxembourg Southeast Asian Studies, University of
Passau (Germany)
Outline
• Objectives
(Geoffrey 1’)• Rationale
– of the paper (Geoffrey 2’)
– of the research in the longer run (Thomas 2’)
• Phnom Penh
(Thomas 2’)• Data
(Thomas 1’)• Methods
(Geoffrey 2’)• Results
(Geoffrey 3’)• Conclusions…so far
(Geoffrey 1’) Σ=14’Objectives
“Roughly” the paper aims at
• Exploratory analysis (ESDA, …mining) of an almost exhaustive field survey dataset of businesses in Phnom Penh
• Exploring current concentration / dispersion + co-location / repulsion patterns
• Identifying and characterising business clusters
• Retail geography should increasingly use the “kitbag of spatial
analysis techniques”
(Birkin et al)Rationale
1. Interdisciplinary work
emerging from face-to-face interactions from random walk within
university corridor with spatial proximity to printers/coffee
constraints
Rationale
1. Interdisciplinary work
2. Long term perspective on evolution of city structure (history and urban economics):
– Emergence of clusters, path-dependence and lock-ins – Inner city agglomeration/dispersion forces
3. Phnom Penh is a particularly appealing laboratory - almost theoretical - setting for business location analysis
– Complete Re-Boot from scratch from 1979 on – Flat plain, almost perfect Manhattan grid network – High transportation costs
– High consumption Demand and free entry for Supply
Imagine a ghost city to be filled in by residents and jobs, see S-GHOST (Peeters et al)
Thomas
• Longer-run rationale
• Phnom Penh
• Data gathering
“pedigree” of the research (concepts & theories)
S o c i a l S c i e n c e s
Traditional Markets Transitional Economies
Colonial & Postcolonial Markets
„Plural Economy“ (J.S. Furnivall, 1944)
„Moral Economy“ (J.C. Scott, 1976)
„Bazaar Economy“ (C. Geertz, 1978)
„Acculturative Economies“ (P. Bohannan &
G. Dalton, 1962)
„Bielefelder Entwicklungssoziologie“
„Informal Economy“
(H.-D. Evers, R. Korff, G. Elwert, 1980s) Development Studies
„Post-Soviet-Transition“
Geography / Economics
„Buzz: face-to-face contact“ (M. Storper &
A.J. Venables, 2004)
„Retail Geography“
„Wirtschaftsgeographie“
(German Economic Geography)
CASE STUDY Phnom Penh
History ??
Forceful eviction of ca. 2 Million city dwellers of Phnom Penh 17 April 1975
ca. 1 Million regular inhabitants 1 Million refugees
The „pavement economy“
of the Chinese shophouse:
Interface zone
The artisan neighbourhood - a `village´ of concrete moulders in the inner city
Subsistence urbanization
- economy of espionage and imitation
- lock-in
vihear – main temple building with buddha statue
squatter in the temple area
pioneer of the business location/business idea
(intern) business locations with `espionage & imitation/variation´
(extern) business locations with `espionage & imitation/variation´
Vat Prayuvong – Squatter and cluster of artisan shops
Dataset to explore
• 14549 businesses
• 558 blocks with at least one shop
• 200 m blocks
• 111 subcategories
• 22 categories
Methods: identify spatial clusters
• Clustering is major objective in geographic data mining
• Abundant literature on identification of subcenters &
spatial clusters
– Density + thresholding
– Polynomial fit on density gradients
– Kernels – GAM
– Spatial scan – …
– LISA
Methods: implemented procedure
For each sub- and category {
For density (x) and % (p) per block { – Mapping of x and p
For 7 distance lags {
• Moran’s I and Scatterplot
• LISA computation
• Reclass of LISA (HH,LL,HL, LH, not signif.)
• Mapping of LISA
• Binary reclass of HH’s
• Counts of HH cluster clumps and clumps size distribution }
– Hierarchical clustering (ward) using HH binaries (Jaccard dissim.) – Hierarchical clustering (ward) for comparison
}
}
library(RColorBrewer)
library(spdep)
library(raster)
Results
• Total business density
• Total business Moran’s
scatterplot
• Moran’s I vs distance lags
– Significant positive global autocorrelation of (almost) all
retail types (robust across 4 to 8 blocks distance)
• Total business LISA clusters
• 1 strong central HH cluster spreading W and S
• Significant LL for newer
residential dvlpt (NW) and river bank (SE)
• No significant HL LH across categories and distance
bands !
Cluster analysis
Our method (Ward, x LISA HH binaries) Standard method (Ward, x)
Cluster analysis
Cluster sum businesses 1 2913
2 4002 3 6043 4 1127 5 249 6 215
Cluster mean n businesses 1 10.59273
2 50.65823 3 46.84496 4 23.00000 5 14.64706 6 23.88889
Conclusions…so far
• Methodological – blocky clusters with no spatial constraints
• Empirical evidences
– Strong evidence of association of businesses of a given type – Resilience of ancient centrality: 1 significant main cluster – 3 significant typology of shops inside main cluster
– Weak periphery but with more diversification
– Espionage, Collusion, Maximize market potential with subtil product differenciation…??
• Next stages
– Additional field work planned to enlighten processes
– Compare with historical maps and prior Pol Pot structure – path-dependence?
Thank you !