Concentration of Manufacturing and Service Sector Activities in Italy: Accounting for Spatial Dependence and Firm Size Distribution

Abstract

De Dominicis L., Arbia G. and De Groot H. L. F. Concentration of manufacturing and service sector activities in Italy: accounting for spatial dependence and firm size distribution, Regional Studies. Empirical analysis of the spatial distribution of economic activity on a discrete space is based on measures that suffer from a series of drawbacks. A methodological advance is proposed here in two respects. First, the analysis is extended to take spatial dependence explicitly into account. Second, differences in the size distribution of firms between territorial units are controlled for. Using data for Italy, exploratory spatial data analysis is applied to identify sectoral location patterns in both the manufacturing industry as well as the business services. It is found that large differences prevail in the geographical concentration of production across sectors.

De Dominicis L., Arbia G. and De Groot H. L. F. 意大利制造业及服务部门产业活动的聚集：考量空间依赖以及公司规模分配，区域研究。 经济活动在离散空间中分布的实证分析存在一系列考量方法上的缺陷。方法论上的改进可以从两方面考虑。首先，分析需要将空间依赖作为考察变量。其次，不同领域单元公司规模分布的差异需要进行控制。使用意大利的数据，解释型空间数据分析可应用于明确制造业以及商业服务的部门区位选择模式。研究发现，跨部门产业的空间聚集存在较大差异。

探索性空间数据分析 公司规模分布 地理积聚 空间自相关 意大利

De Dominicis L., Arbia G. et De Groot H. L. F. La concentration de l'industrie et du secteur tertiaire en Italie: tenir compte de la dépendance géographique et de la distribution des entreprises suivant leur taille, Regional Studies. L'analyse empirique de la distribution géographique de l'activité économique dans une zone bien délimitée est fondée sur des mesures qui souffrent d'une série d'inconvénients. On propose une avance méthodologique à deux égards. Primo, l'analyse se voit approfondir afin de tenir explicitement compte de la dépendance géographique. Secundo, on tient compte des différences de la distribution de la taille des entreprises sur le plan interrégional. Employant des données pour l'Italie, on applique une analyse géographique exploratoire des données pour identifier la distribution des emplacements sectoriels à la fois dans l'industrie ainsi que dans les services aux entreprises. Il s'avère que de grandes différences persistent pour ce qui est de la concentration de la production à travers les secteurs.

Analyse géographique exploratoire des données  Distrubtion de la taille des entreprises  Concentration géographique Autocorrélation géographique Italie

De Dominicis L., Arbia G. und De Groot H. L. F. Konzentration von Aktivitäten des produzierenden und Dienstleistungssektors in Italien: Berücksichtigung von räumlicher Dependenz und Firmengrößenverteilung, Regional Studies. Die empirische Analyse der räumlichen Verteilung von Wirtschaftsaktivität auf einen eigenständigen Raum beruht auf Maßstäben, die mit einer Reihe von Nachteilen verbunden sind. In diesem Beitrag wird in zweierlei Hinsicht eine methodologische Weiterentwicklung vorgestellt. Erstens wird die Analyse unter expliziter Berücksichtigung der räumlichen Abhängigkeit erweitert. Zweitens wird auf Unterschiede hinsichtlich der Größenverteilung von Firmen in verschiedenen territorialen Einheiten kontrolliert. Anhand von Daten für Italien wird die explorative räumliche Datenanalyse zur Identifizierung von sektoralen Standortsmustern für die produzierende Industrie sowie für Geschäftsdienste angewandt. Hinsichtlich der geografischen Konzentration der Produktion in verschiedenen Sektoren werden erhebliche Unterschiede festgestellt.

Explorative räumliche Datenanalyse Firmengrößenverteilung Geografische Konzentration Räumliche Autokorrelation Italien

De Dominicis L., Arbia G. y De Groot H. L. F. Concentración de las actividades del sector manufacturero y de servicios en Italia: considerando la dependencia espacial y la distribución del tamaño de las empresas, Regional Studies. Los análisis empíricos de la distribución espacial de la actividad económica en un espacio diferenciado se basan en medidas relacionadas con una serie de inconvenientes. Aquí proponemos un avance metodológico desde dos perspectivas. Primero, ampliamos el análisis para tener en cuenta específicamente la dependencia espacial. Segundo, controlamos las diferencias en la distribución del tamaño de las empresas entre las unidades territoriales. A partir de datos para Italia, aplicamos un análisis exploratorio de los datos espaciales, lo que nos permite identificar los modelos de ubicación sectorial tanto en la industria manufacturera como en los servicios comerciales. Observamos que se imponen grandes diferencias en la concentración geográfica de la producción en los diferentes sectores.

Análisis exploratorio de datos espaciales Distribución del tamaño de las empresas Concentración geográfica Autocorrelación espacial Italia

Keywords:

• Exploratory spatial data analysis
• Firm size distribution
• Geographic concentration
• Spatial autocorrelation
• Italy

JEL classifications:

• C12
• R12
• R30

Acknowledgements

The authors thank the Editor and three anonymous referees for insightful comments. The usual disclaimer applies. The first author acknowledges financial support from the European Union's Sixth Framework Programme for Research and Technological Development. The views expressed are purely those of the author and cannot in any circumstances be regarded as stating an official position of the European Commission.

Notes

For a good summary of the main results, see Combes and Overman (2004).

Some other authors employ a rather different approach, which is not based on the standard concentration measures, but aims at studying the spatial clustering of activities in a continuous space. Such an approach was introduced by Arbia and Espa (1996) and subsequently exploited by, amongst others, Marcon and Puech (2003) for France, Quah and Simpson (2003) and Duranton and Overman (2005) for the clustering of industries in the UK, and by Arbia et al. (2008) for the co-agglomeration of high-technology industries in Italy.

NACE is the acronym for Nomenclature générale des activités économiques dans les Communautés Européennes.

An Annex with the complete set of the sectors in the data set is available from the authors upon request.

Starting from a data set at the municipal level, the data are aggregated into larger administrative units (more specifically, Nomenclature des Unités Territoriales Statistiques (NUTS)-3 and NUTS-2 regions) and into functional regions (local labour market areas, or LLMAs). Results for the NUTS regions are available from the authors upon request.

The European Commission has recently recognized the central role of the LLSs. Following a period of negotiation between the European authorities and the Italian government, the LLSs have become the territorial units used by the European Union to identify the areas eligible under Objective 2 in the Northern and Central regions of Italy for the 2000–2006 programming period (Commission Decision 2000/530/EC of the 27 July 2000, listing the areas of Italy eligible under Objective 2 of the Structural Funds for the period 2000–2006; available at: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2000:223:0001:0069:EN:PDF).

The authors are aware that the choice of spatial units might influence some of the results on agglomeration since the definition of LLS is based on daily commuting flows. A sector is more likely to be found agglomerated if it is largely dominant in a dense area of commuting. However, the authors believe that for the purpose of this study they are still the most appropriate geographical units as they rely on real economic features. They reflect, as closely as possible, commuting fields of what can be considered a local economy and they are far less arbitrary than administrative units (O'Donoghue and Gleave, 2004).

For details on the derivation of the measure of geographical concentration, see Ellison and Glaeser (1997).

Ellison and Glaeser show that the expected value of this measure is zero if plants are randomly located, with any positive value of the index interpreted as localization. In particular, values between zero and 0.02 are interpreted as weak localization, and anything above 0.05 as a strong tendency to localize (Ellison and Glaeser, 1997).

As shown by Ellison and Glaeser (1997) and Maurel and Sedillot (1999), the variance of the estimator under the null hypothesis of no spillovers γ = 0 is given by:

The result can be used to perform a t-test comparing the value of the index with twice its standard deviation, which, under the assumption of normality, is a test at the 5% significance level. Significant values of the test indicate that the observed degree of concentration deviates significantly from a situation of random location of the firms.

An Annex with the full set of results is available from the authors upon request.

Results are presented in this paper only for functional regions. However, the measure for Italian provinces (NUTS-3) and regions (NUTS-2) was computed. For the full set of results, see the Annex available from the quthors upon request.

However, in the last years, thanks to the financial and fiscal incentives available to the Objective 1 regions, FIAT has decentralized part of its production to the southern regions of Italy.

The concept of the ‘Third Italy’ was introduced in the late 1970s. At that time it became apparent that little economic progress was realized in the South (‘Second Italy’), and that the traditionally rich Northwest (‘First Italy’) was facing a deep crisis, while in contrast the Northeast and Centre of Italy showed fast growth.

By convention, the diagonal elements of the weights matrix are set to zero.

The robustness of the results reported in the paper were checked. More specifically, the exercise was repeated using spatial weights matrices based on the squared of the inverse distance between pairs of locations (with different threshold levels), and on first-order contiguity. The Annex available online contains the results of the sensitivity analysis performed on the different spatial weight matrices.

Inference is based on 999 permutations at the 0.05 significance level.

Amongst others, applications are found of local indicators of spatial association for the analysis of the distribution of regional income and structural funds in Europe (Dall'Erba, 2005), and for the study of local agglomeration patterns in Paris and its surroundings (Guillain and Le Gallo, 2010).

The Annex available online contains the results for all sectors.

The traditional sectors consist of the low-technology manufacturing and the medium- to low-technology manufacturing.

In Italy the district model consists of a high concentration of small firms specialized in the production of a specific item and grouped together in the same territorial area. For additional information on industrial districts in Italy and for an overview of their history and main production activities, see http://www.distretti.org/.

Sectors 16–21, 23, 25–30, 33–37, 52, 55, 60, 61, 63, 64, 66, 73 and 71.

Sectors 15, 22, 24, 31, 32, 40, 41, 45, 50, 51, 62, 65, 67, 70, 72 and 74.