Urban Centrality and retail’s revenues
If the stores located in the center of the city could generate higher sales volumes than the ones in the periphery?
We introduce the centrality indicators to measure the centrality of spatial networks.
Betweenness centrality and network size
The betweenness centrality was computed to measure the local accessibility in the spatial networks. By examining the subset of nodes and edges in the network, Node i’s betweenness was determined to consider the network distance d; that is, the network distance d was changed from a neighborhood scale to a city scale for node i, and a betweenness for every node i computed.
Results of betweenness scores for the different network size
It shows the different network distances d for node i. Changing d to compute the betweenness for node i results in different scores for node i. In the smaller-scale network, the locations with higher betweenness scores were located on local arteries that connected public spaces at the neighborhood scale. The more accessible locations gradually changed as the network size increased and became more coincident with the large-scale inter-district avenues for the largest value of d.
Representative distribution of nodes’ betweenness scores in the street network and its change by the network size
As different dependence patterns of betweenness as a function of d were found, to quantitatively identify the particular clusters. We appied the k-means clustering algorithm to classify the streets into four clusters.
The nodes in the first and second cluster were quite evenly distributed over the city, and the nodes in the third (locally important) and fourth clusters (important at city level) had clear spatial signatures. The third cluster, which was the locally important nodes, were exclusively located in the historic center and corresponded to the arteries constructed at the beginning of the city formation. However, the nodes in the fourth cluster corresponded to the larger avenues that connected the different parts of the city.
It shows the normalized revenue for stores located at a distance (x-axis) from the road network nodes belonging to cluster 3 (locally important nodes), from which it can be seen that the normalized store revenue pattern was higher for stores near the locally accessible nodes.
However, the positive effect of city-scale accessibility on retail revenue was apparent. All shop categories had a peak in normalized revenue for shorter distances to the city-scale accessible nodes; i.e., nodes belonging to cluster 4. It was also found that the revenue for the stores located near the accessible nodes at either the local (Figure 4(b)) or the city scale (Figure 4(d)) was much higher than average.
The locally accessible nodes (cluster 3) contributed positively to daily-use stores revenue. Customers of daily-use stores (e.g., butchers, fishmongers) would be more likely to be drawn from nearby locations rather than from far away; that is, locally accessible shops are more likely to observe more customers. However, locally accessible nodes were not found to have the same effect for other categories (e.g., B2 and B3) because the customers tended to come from a much wider district or city scale areas.
As the nodes in cluster 4 were found to increase store revenue independent of type, these nodes could be considered central places when considering larger network sizes, such as district or city scales as they more frequently captured passers-by from the whole city than other places. Therefore, these locations are advantageous for non-daily-use category stores (e.g., department stores) that derive their customers from a wider area than the local neighborhood.
Yoshimura, Y., Santi, P., Murillo Arias, J., Zheng, S., Ratti, C (2020). Spatial clustering: Influence of urban street networks on retail sales volumes, Environment and Planning B: Urban Analytics and City Science (accepted).