Urban size and KIBS vertical disintegration: the case of Milan

ABSTRACT

The literature shows that market size favours firms’ vertical disintegration because it reduces transaction and knowledge coordination costs. The existing evidence is based primarily on cross-sectional data and on the manufacturing sector, with little or no attention is given to longitudinal data and services. This paper estimates the effect of urban size on the purchase of service inputs by knowledge-intensive business service firms located in the metropolitan region of Milan. Relying on a rich panel dataset, and adopting different econometric techniques to overcome endogeneity, we find a positive and statistically significant effect of increased urban population on the degree of KIBS vertical disintegration only when urban size is measured at the province level. We also find that this effect holds particularly in the case of traditional professional- rather than technology-related activities.

KEYWORDS:

• KIBS
• panel data
• urban size
• vertical disintegration

JEL:

• L84
• R12
• R34

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Roberto Antonietti http://orcid.org/0000-0002-2172-4062

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

Notes

1. Using the Scopus database by Elsevier, a simple search of the articles, books, book chapters and reviews published in social science and humanities with the words ‘knowledge-intensive business services’ in the title, abstract or keywords reveals that the number of documents was 4 until 1996, 80 until 2006, rising up to a 537 in 2016 (last search 25 July 2016). When ‘vertical integration’, ‘vertical disintegration’ or ‘division of labor’ is added in the search engine, only three to five documents appear.

2. Since we are interested in local provision of intermediate inputs, and taking account of the nature of the firm being analysed, we explicitly omit purchase of material inputs, which have no local dimension in provision.

3. We decided to use age as a time-invariant variable because, once we put the model into first differences, it would give only values of 1.

4. We tested whether the difference in the estimated coefficients of ln(POP)_k between P-KIBS and T-KIBS is statistically significant through a test of equality of coefficients across regressions. The tests always reject the null hypothesis of equality of coefficients at the 1% level.

5. The low R² is consistent with that found in previous works on industry concentration and vertical disintegration (Holmes, 1999; Li & Lu, 2009) or in works on agglomeration externalities using panel data (Martin et al., 2011).