Mapping industrial patterns in spatial agglomeration: A SOM approach to Italian industrial districts

Abstract

The paper presents a new approach based on Self-Organizing Maps (SOM) and a new index called Relative Industrial Relevance (RIR) to discover, track and analyze spatial agglomeration of economic activities. By comparing patterns of local employment, this methodology shows how the local supply of human capital can explain the advantages generating spatial agglomerations. The reference case for this research is Italy, which has developed one of the most remarkable and studied example of spatial agglomerations, the Industrial Districts (IDs). IDs are traditionally identified by indexes which measure the physical concentration of firms belonging to a given industry, but are unable to seize the overall productive structure of the local economy. Employing the Italian Clothing Industry as test bed, the approach proposed in this paper identifies spatial agglomerations in terms of industry patterns and not of industry concentration. This methodology can offer a new basis to analyze the multiple pattern of local development.

Measuring spatial concentration: A transportation problem approach

“…The MAUP was again discussed by Lafourcade and Mion (), and, more recently, Kopczewska () provided an in‐depth analysis of the cluster‐based measures of the geographical and sectoral concentration offered in the regional science literature, in which the MAUP had been emphasized. Other issues related to the limits in identifying spatial agglomeration through raw concentration measures have been discussed by Carlei and Nuccio (). And Chain, Santos, Castro, and Prado () have provided a bibliometric analysis of the methods used for the measurement of industrial clusters.…”

SPAG: Index of spatial agglomeration

“…The above criteria were challenging mostly for cluster‐based measures and evoked the progress in literature: the significance test applied to traditional measures as a bootstrap test for LQ (Tian, ) or adding spatial components to them (Arbia & Piras, ; Bickenbach & Bode, ; Carlei & Nuccio, ; Guillain & Le Gallo, ; Guimaraes et al, ; Sohn, ). Distance‐based measures fulfil most of the criteria specified by Duranton and Overman ().…”

“…Additionally they do not see the heterogeneity inside the region, as they compare regions and sectors between themselves as homogenous units. Improved cluster‐based indices, including the spatial structure expressed usually by the spatial weights matrix W (Arbia, ; Arbia & Piras, ; Bickenbach & Bode, ; Carlei & Nuccio, ; Guillain & Le Gallo, ; Guimaraes, Figueiredo, & Woodward, ; Sohn, ), are in the middle‐way as they add information on spatial autocorrelation, but keep the minorities of the a‐spatial indices.…”

Evaluation of Land Intensive Use in Shanghai Pilot Free Trade Zone

“…Other commonly used methods include the limit condition method, fuzzy clustering analysis, principal component analysis (PCA), hierarchical analysis, multi-factor comprehensive evaluation, evaluation methods based on the pressure, state and response (PSR) model or the remote sensing images, stochastic frontier analysis, genetic algorithm, and BP neural network, etc. [32][33][34][35][36][37][38][39][40][41][42][43][44][45]. Based on the concept of sustainable intensification and ecological constraints, Qian et al developed a sustainable intensification variable model to conduct a moderately intensive land use evaluation in Jinan City, China [46].…”