Does spatial proximity matter? Micro- evidence from Italy

Using a similar empirical framework, Henderson et al.­(1995)­find­that­localisation plays a positive role in mature capital-goods sectors, while differentiation of the productive structure (variety), which should generate cross-fertilisation of ideas between different industries, has a positive impact only in the case of high tech­ industries.­Using­French­data,­Combes­(2000)­also­finds­a­ rather­negative­ impact of specialisation on employment growth in both the industry and service sectors. Finally, Forni and Paba (2002), using information on a cross-section of 995­Italian­LLSs­for­the­period­1971-1991­find­that­in­most­cases­specialisation­ and variety positively affect growth, but the variety is different for each industry. Moreover, they note that, consistent with Marshall (1920), in order to capture the­ spillover-­generating­process­a­ size­effect­needs­ to­be­added­ to­ the­ specialisation effect. Glaeser et al.’s­(1992)­approach­has­been­replicated­in­the­contexts­of­different countries in order to provide further evidence on these issues. Nonetheless, the­various­results­obtained­from­empirical­research­in­this­field­are­controversial­ such that currently there is not a unique model explaining the link between employment growth and the structure of the local economy. In particular, some studies­referring­to­the­Italian­case­find­that­specialisation­has­a­negative­impact­ on local growth, while diversity plays a positive role (see, among others, Cainelli and Leoncini 1999; Cainelli et al. 2001; Cunat and Peri 2001; Usai and Paci 2003; Paci and Usai 2006; Mameli et al. 2007). This empirical literature was extended by several studies that analyse the impact of measures of agglomeration economies both on employment growth (as in­the­original­body­of­literature­referred­to),­and­on­productivity­or­firms’­total­ factor productivity (TFP) growth (de Lucio et al. 2002; Henderson 2003; Cingano and Schivardi 2004; Martin et al.­2008).­The­findings­within­this­new­ strand­of­empirical­ research­are­also­rather­puzzling.­For­example,­de­Lucio et al. (2002) investigating the relationship between labour productivity and spatial agglomeration­at­the­level­of­the­50­Spanish­provinces­for­the­period­1978-1992,­ find­that­variety­plays­a­role­in­labour­productivity­growth,­and­find­a­U-­shaped­ effect for specialisation. According to their results, low levels of specialisation reduce productivity growth, and high levels foster it. In contrast, Cingano and Schivardi­ (2004),­ using­ firm-­level­ based­ TFP­ indicators,­ show­ that­ specialisation, calculated at the level of the 784 Italian LLSs, has a positive impact on firm­productivity­growth,­but­that­variety­has­no­significant­effect.­Taking­local­ employment growth as the dependent variable, Cingano and Schivardi (2004) show that the specialisation effect is reversed and becomes negative, while variety­has­a­significant­and­positive­ impact­on­employment­growth,­ thus­confirming­Glaeser­et al.’s­results.­On­the­other­hand,­Henderson­(2003),­using­the­ Longitudinal­ Research­ Database­ (LRD)­ of­ the­ US­ Census­ Bureau,­ finds­ that­ localisation economies have strong positive effects on productivity at plant level in­high­tech­industries,­but­not­in­machinery­industries,­and­finds­little­evidence­ of urbanisation economies. Finally, Martin et al. (2008), using French individual firm­data­from­1996­to­2004,­find­no­significant­effect­of­spatial­agglomeration­ on­ firm­ productivity.­More­ precisely,­ they­ find­ that­ French­ firms­ benefit­ from­

localisation,­ but­ not­ from­urbanisation­ economies.­However,­ the­benefits­ from­ industrial­clustering­–­even­if­highly­significant­from­a­statistical­point­of­view­–­ are quite modest in terms of magnitude. The use of TFP measures is an obvious and notable improvement to these studies which, however, must also acknowledge some of the drawbacks related to other measurement and empirical issues: for example, the use of sample of plant data (Henderson 2003; Martin et al. 2008) and the problems in the case of Cingano­and­Schivardi’s­paper­of­sample­selection. ­ Another­major­ shortcoming­ of­ all­ these­ studies­ is­ that­ they­ refer­ to­ exogenously­defined­geographic­units­such­as­SMAs,­LLSs,­or­administrative­regions­ or­provinces.­A­consequence­of­this­choice­is­that­all­these­studies­find­it­difficult­ to deal with a rather relevant aspect of these phenomena: namely, the attenuation of agglomeration economies over space. In the recent literature two different approaches­were­proposed­to­deal­with­this­issue.­First,­Desmet­and­Fafchamps­ (2005), using US county data for 1972 and 2000, try to overcome it by assuming that­a­county’s­employment­growth­is­not­only­affected­by­the­county­under­consideration,­but­also­by­all­“near”­counties.­Van­Oort­(2007)­also­ tries­ to­ tackle­ this issue by considering spatial dependence, though his study has some problems with the robustness of his results. A second approach was developed by Wallsten (2001). Calculating by means of a GIS program the distance between each­firm-­pair,­ he­ investigates­ spatial­ spillovers­ at­ the­firm­ level­ over­ discrete­ distances. Finally, Rosenthal and Strange (2008), using 2000 census data to estimate the relationship of agglomeration and proximity to human capital to wages­and­taking­a­geographic­approach,­find­that­the­benefit­of­spatial­concentration tends to attenuate sharply with distance.