Distance-based agglomeration externalities and neighbouring firms’ characteristics

ABSTRACT

This paper tests the hypothesis that firms with different characteristics can differ in their capability to produce local externalities by investigating the relationship between firm-specific distance-based weighted agglomeration measures and firms’ short-run productivity growth in the Italian manufacturing industry. The results suggest that positive localization economies increase with distance when neighbouring firms’ characteristics are accounted for. Diversification-type forces have negative effects on productivity growth at short distances, while there are positive effects at longer distances regardless of the weighting scheme considered. Moreover, the negative effect of inter-industry externalities seems to persist over distance when neighbouring firms’ characteristics are accounted for.

KEYWORDS:

• distance-based agglomeration
• neighbourhood characteristics
• total factor productivity
• manufacturing
• Italy

JEL:

• C3
• D24
• R12

ACKNOWLEDGEMENTS

The authors thank the editor, Thomas Kemeny, and two anonymous reviewers for numerous suggestions that helped to improve this paper. The authors also acknowledge Eric Marcon (AgroParisTech) and Steve Gibbons (LSE) for valuable advice about the computation of the distance-based agglomeration measures, as well as the participants at the 55th Congress of the European Regional Science Association, Lisbon, Portugal, August 2015, for useful comments on a previous version of the paper. The usual disclaimers apply.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

ORCiD

Giulio Cainelli http://orcid.org/0000-0003-0902-1847

Notes

1. The MAUP has been widely investigated by statisticians and quantitative geographers (e.g., Arbia, 1989; Amrhein, 1995; Wong & Amrhein, 1996).

2. The literature proposes alternative solutions to mitigate the MAUP. Some contributions suggest controlling for extra-region spillovers through spatially lagged agglomeration variables computed within administrative or labour market regions (e.g., Burger et al., 2010; Van Oort, 2007). Others propose a multilevel approach to enable simultaneous modelling at micro- and macro-levels of analysis (e.g., Van Oort, Burger, Knoben, & Raspe, 2012).

3. The reviewed works extend to the measurement of agglomeration economies the micro-geographical approaches used to identify the geographical concentration of economic activities through spatial statistics (probability or cumulative density functions) which use pair distances between firms to evaluate at which geographical scale a particular industry shows a clustering pattern (e.g., Arbia & Espa, 1996; Duranton & Overman, 2005; Marcon & Puech, 2010; Scholl & Brenner, 2016).

4. The main reason for this choice is lack of georeferenced micro-data at the census level. We are conscious that this represents a limit of our analysis. Cainelli and Lupi (2010) and Gabriele et al. (2013) adopt a similar approach for the Italian case. Similarly, Martin et al. (2011) construct agglomeration measures using sample data drawn from the French Annual Business Survey, a dataset covering all firms with more than 20 employees. The advantage of our dataset is that it also covers firms with fewer than 20 employees, which represent the majority in our sample. In our opinion, this degree of coverage is particularly important in the context of spatial agglomeration.

5. Distance-based agglomeration measures are firm specific, them being centred at each single firm. The centroid of each firm corresponds to the geographical coordinates obtained from the firm’s exact address. This means that each firm, for which the exact address was available, has been geocoded. Although this approach provides great precision in the identification of a firm’s neighbourhood, it may also present disadvantages if geographical coordinates refer to the headquarter rather than to (productive or commercial) plants of a multi-plant firm. As the AIDA databank provides information on the headquarters, this issue may represent a drawback of this study. However, the nature and characteristics of the Italian industrial structure limit this drawback, it being driven by many small firms such that the headquarter tends to coincide with the principal (or unique) operative plant. For instance, according to the 2009 ASIA Archive provided by the Istituto Nazionale di Statistica (ISTAT), multi-plant firms represent only 10.94% of active firms operating in the manufacturing industries covered in this analysis.

6. The agglomeration variables in equation (1) are computed using the R Project for Statistical Computing. Original coding is based on the ‘dbmss’ R package developed by Marcon, Lang, Traissac, and Puech (2012).

7. Rosenthal and Strange (2003), Gabriele et al. (2013) and Deltas et al. (2015), among others, capture diversification externalities simply considering industries different from the reference one.