Skip to main content

Local manufacturing productivity markers: an empirical study of the Italian counties


We analyse productivity differentials across about 63,000 manufacturing firms located in 103 Italian counties, in order to shed light on the relation between the business environment and firm performance. We find that a limited set of local variables related to institutional quality, local credit development, market access and innovation environment significantly contribute to explaining manufacturing productivity differentials in Italy. Our empirical findings confirm that firm competitiveness reacts to the local business environment on a multidimensional scale. This suggests that better targeted regional policies at the national and EU level, including measures for fostering convergence or decentralizing wage negotiations, should take into account the interdependence between productivity and the economic environment.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3


  1. See for instance Melitz and Ottaviano (2008).

  2. Tabellini (2010) suggests, for example, that the judicial system performs differently in Southern and Northern Italy, with judges taking much longer to complete investigations and to rule on civil cases in the South than in the North, even though the formal framework is similar.

  3. Please note that we estimate Eq. (1) separately for each 2-digit (NACE rev3) sector considered in the paper.

  4. TL reduces to a CD once you drop squared and the interaction terms.

  5. As an example, the labour elasticity implied by the TL production function is equal to \({\beta }_{{ it}}^{l} =\beta _{l} +2\beta _{ll} {l}_{{ it}} +\beta _{lk} {k}_{{it}}\).

  6. Moreover, the GMM approach is also robust to Ackerberg et al. (2006) critique that labour coefficient will not be identified in the two step estimation if its choice is a function of unobserved productivity shock, see Wooldridge (2009).

  7. Quintile dummies are computed by sector and year and guarantee a larger deal of flexibility, we also use the log of sales as control for production size and results are robust and available upon request from the authors.

  8. NUTS stand for the Nomenclature of Territorial Units for Statistics, a hierarchical system used by Eurostat to produce regional statistics. The last revision of the NUTS classification (2013) divides European Union territory in 98 major socio-economic regions (NUTS1), 276 basic regions for the application of regional policies (NUTS2) and 1342 small regions for specific diagnoses (NUTS3). In the rest of the paper, we focus on the 103 (NUTS3) counties defining the Italian territory.

  9. See Syverson (2011) for a detailed discussion on both internal and external markers of firm productivity. Note that county’s covariates are lagged to limit simultaneity bias; reverse causality is ruled out by the assumption that an individual firm is not able to fully determine local business environment.

  10. For employment size, we identify 4 classes: less than 10 workers, from 10 to 49, from 50 to 249, above 250.

  11. We choose to let errors to be correlated at the regional level (NUTS2) a higher geographical scale than counties (NUTS 3) to account for the fact that one of the covariates—R&D expenditure—is only available at the NUTS2 detail.

  12. Unfortunately R&D expenditure is only available at the regional level (NUTS2) not at the county one (NUTS3).

  13. The census is carried out by the Italian National Institute of Statistics (ISTAT).

  14. In counties with municipalities submitted to mandatory administration public spending contract of about 0.5% of county’s value added.

  15. The number of municipalities under mandatory administration is normalized by the overall number of municipalities under the county administration. This information is publicly available from the Italian Ministry of Internal Affairs. Since this event is unlikely to be correlated with firm level innovations, we include the contemporaneous value in order to maximize sample size.

  16. Unfortunately information on outstanding credit at the county level is available only from 1998.

  17. Source: Bank of Italy and the Italian National Institute of Statistics (ISTAT).

  18. The European Observation Network for Territorial Development and Cohesion adopted by the European Commission on 7 November 2007 ( The accessibility index is equal to 1 for the average European NUTS 3 territory and increases for better connected counties.

  19. As a baseline, we define an LLMA as UA if the population is above the 90th percentile of the distribution in year 2001. In Table 9 in “Appendix”, however, we report the estimation results for different thresholds: above 75th, 95th, 99th percentile as well as the threshold of 500 thousand residents used in Giacinto et al. (2014); our main findings are fully robust to the UA cut-off.

  20. A possible concern using those data is that it tends to under-represent small and medium enterprises. In terms of number of firms, indeed, AIDA reports 78.5% of firms below with less than 50 workers for year 2010, while those represent around 95% of the sample in the business register.

  21. More specifically, we use 2-digit production prices to deflate value added, total fixed assets prices for capital, production prices of intermediate inputs for materials, setting 2000 as the base year. Note that since we only observe the book value of capital, we build real capital series using a perpetual inventory method. For a general firm i, the real value of capital stock at time t is derived as the real stock at time \(t-1\) minus depreciation plus the real investment between \(t-1\) and t. Depreciation rates are at the 2-digit industry level and are measured as the ratio between aggregate amortizations and net capital from the national institute of statistics (ISTAT).

  22. Extreme values may be due to miss reporting but may also be induced by management strategies (e.g. by moving all the real estate assets to a separate legal entity) that may affect production function estimates. Results are reported in Table 10 and fully confirm our main findings.

  23. Following Giacinto et al. (2014) in the OLS specification, we also include three-digit industry fixed effects to allow for more heterogeneity at the industry level.

  24. Note that the reported coefficients for the TL specification are industry averages.

  25. Note that the 0.90 return to scale parameter is slightly below constant return to scale but it is significantly higher than the one in Gopinath et al. (2015), where they find 0.71 for Italy, and 0.80 for the overall sample (France, Germany, Italy, Norway, Portugal, and Spain), using the Amadeus dataset (covering all Europe) by Bureau van Dijk, the same data producer of Aida (covering only Italy).

  26. The TFP (GMM-TL) has been centred to a zero mean by sector in order to control for differences in industrial composition between the North and the South of the country. The difference in mean is statistically significant at 1% confidence level.

  27. Recall that TFP is measured in natural logs.

  28. As a further robustness check in Table 8 in Annex, we report a number of nested model specifications where the individual regressors are added one at a time in order to check the stability of regression coefficients as the model specification is progressively enlarged. Results are largely confirmed both for the sign and for the significance of the individual variables.

  29. Where \(X^{*}\) is derived as \(X^{*}=\frac{\beta _{\mathrm{age}} }{-2*\beta _{\mathrm{age}^{2}}}\). Since the underlying variable is expressed in logarithm, the corresponding levels are given by: \(e^{3.05}=21.11\)

  30. They show that both geographical distance and time travel to next urban agglomeration are negatively associated with regional per capita GDP growth, but travel time seems to be a stronger predictor of regional growth.

  31. In particular, Cingano and Schivardi (2005) find that firm productivity growth is positively affected by both localization economies and overall city size.


  • Acconcia A, Corsetti G, Simonelli S (2014) Mafia and public spending: evidence on the fiscal multiplier from a quasi-experiment. Am Econ Rev 104(7):2185–2209

    Article  Google Scholar 

  • Acemoglu D, Dell M (2010) Productivity Differences between and within countries. Am Econ J Macroecon Am Econ Assoc 2(1):169–188

    Article  Google Scholar 

  • Ackerberg D, Caves K, Frazer G (2006) Structural identification of production functions. MPRA Paper 38349, University Library of Munich

  • Ahrend R, Schumann A (2014) Does regional economic growth depend on proximity to urban centres? OECD Regional Development Working Papers, 2014/07, OECD Publishing

  • Aiello F, Scoppa V (2000) Uneven regional development in Italy: explaining differences in productivity levels. Giornale degli Economisti 59(2):270–298

    Google Scholar 

  • Ascari G, di Cosmo V (2005) Determinants of total factor productivity in the Italian regions. Sci Reg 4(2):27–49

    Google Scholar 

  • Beck T, Demirguc-Kunt A, Laeven L, Levine R (2008) Finance firm size, and growth. J Money Credit Bank 40(7):1379–1405

    Article  Google Scholar 

  • Breinlich H, Ottaviano GIP, Temple JRW (2014) Regional growth and regional decline, chapter 4. In: Aghion P, Durlauf SN (eds) Handbook of economic growth, vol 2. Elsevier, Amsterdam, pp 683–779

  • Bronzini R, Piselli P (2009) Determinants of long-run regional productivity with geographical spillovers: the role of R&D, human capital and public infrastructure. Reg Sci Urban Econ 39(2):187–199

    Article  Google Scholar 

  • Cameron AC, Miller DL (2015) A practitioner’s guide to cluster-robust inference. J Hum Resour 50(2):317–373

    Article  Google Scholar 

  • Cannari L, Magnani M, Pellegrini G (2009) Quali politiche per il Sud? Il ruolo delle politiche nazionali e regionali nell’ultimo decennio. Bank of Italy Occasional Papers, no. 50. Rome

  • Caselli F (2005) Accounting for cross-country income differences, handbook of economic growth, chapter 9. In: Aghion P, Durlauf S (eds) Handbook of economic growth, vol 1, 1st edn. Elsevier, Amsterdam, pp 679–741

    Google Scholar 

  • Castelli A, Dwyer GP, Hasan I (2009) Bank relationships and firms’ financial performance: the Italian experience, Discussion Papers 36. Bank of Finland Research

  • Cingano F, Schivardi F (2005) Identifying the sources of local productivity growth. In: Banca d’Italia, Local Economies and Internationalization in Italy, Rome

  • Combes PP, Duranton G, Gobillon L, Puga D, Roux S (2012) The productivity advantages of large cities: distinguishing agglomeration from firm selection. Econometrica 80(6):2543–2594

  • De Loecker J, Goldberg PK (2014) Firm performance in a global market. Annu Rev Econ 6(1):201–227

    Article  Google Scholar 

  • Di Giacinto V, Gomellini M, Micucci G, Pagnini M (2014) Mapping local productivity advantages in Italy: industrial districts, cities or both? J Econ Geogr 14(2):365–394

    Article  Google Scholar 

  • Eickelpasch A, Lejpras A, Stephan A (2007) Hard and soft locational factors, innovativeness and firm performance—an empirical test of porter’s diamond model at the micro-level, CESIS. Electronic Working Paper Series, Paper no. 109

  • Escribano A, Guasch JL, de Orte M, Pena J (2008) Investment climate assessment based on demean Olley and Pakes decompositions: methodology and application to Turkey’s investment climate survey, Universidad Carlos III de Madrid, Working Paper n. 20, Economic Series

  • Fox JT, Smeets V (2011) Does input quality drive measured differences in firm productivity? National Bureau of Economic Research, Working Paper n. 16853

  • Gopinath G, Kalemli-Ozcan S, Karabarbounis L, Villegas-Sanchez C (2015) Capital allocation and productivity in South Europe, NBER Working Papers 21453, National Bureau of Economic Research, Inc

  • Guiso L (2003) Small business finance in Italy. EIB Papers 10/2003, European Investment Bank—Economics Department

  • Guiso L, Sapienza P, Zingales LM (2004) Does local financial development matter? Q J Econ 119:929–69

    Article  Google Scholar 

  • Hsieh CT, Klenow PJ (2009) Misallocation and manufacturing TFP in China and India. Q J Econ 124(4):1403–48

    Article  Google Scholar 

  • Hsieh CT, Klenow PJ (2010) Development accounting. Am Econ J Macroecon Am Econ Assoc 2(1):207–23

    Article  Google Scholar 

  • Ilmakunnas P, Maliranta M, Vainiomäki J (2004) The roles of employer and employee characteristics for plant productivity. J Product Anal 21(3):249–276

    Article  Google Scholar 

  • Jorgenson DW, Mun SH, Stiroh KJ (2005) Productivity, information technology and the American Growth Resurgence, vol 3. MIT Press, Cambridge

    Google Scholar 

  • King R, Levine R (1993) Finance, entrepreneurship, and growth: theory and evidence. J Monet Econ 32:513–542

    Article  Google Scholar 

  • Laeven L, Levine R, Michalopoulos S (2015) Financial innovation and endogenous growth. J Finance Intermed 24(1):1–24

    Article  Google Scholar 

  • Levine R (2005) Finance and growth: theory and evidence. In: Aghion P, Durlauf SN (eds) Handbook of economic growth. Elsevier, North-Holland, New York, pp 866–934

    Google Scholar 

  • Levine R (2006) Finance and growth: theory and evidence. In: Aghion P, Durlauf S (eds) Handbook of economic growth. Elsevier, North-Holland, New York, pp 865–934

    Google Scholar 

  • Levinshon J, Petrin A (2003) Estimating production functions using inputs to control for unobservables. Rev Econ Stud 70(2):317–341

    Article  Google Scholar 

  • Medda G, Piga C (2014) Technological spillovers and productivity in Italian manufacturing firms. J Prod Anal 41(3):419–434

    Article  Google Scholar 

  • Melitz M, Ottaviano G (2008) Market size, trade and productivity. Rev Econ Stud 75:295–316

    Article  Google Scholar 

  • Oliner SD, Sichel DE, Stiroh JS (2007) Explaining a productive decade. In: Brookings papers on economic activity, vol 1, pp 81–137

  • Syverson C (2011) What determines productivity? J Econ Lit 49(2):326–365

    Article  Google Scholar 

  • Tabellini G (2010) Culture and institutions: economic development in the regions of Europe. J Eur Econ Assoc 8(4):677–716

    Article  Google Scholar 

  • Wooldridge JM (2009) On estimating firm-level production functions using proxy variables to control for unobservables. Econ Lett 104(3):112–114

    Article  Google Scholar 

Download references


We would like to acknowledge comments from Tullio Buccellato, Maurizio Fiaschetti, Fabiano Schivardi, participants in conferences held at LUISS, Rome and the Sant’Anna School in Pisa, as well as two anonymous referees.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Stefano Manzocchi.



In Table 8, we report the estimate for nested model specifications where the individual regressors are added one at a time in order to check the stability of regression coefficients as the model specification is progressively enlarged. Estimated parameters are stable, both in magnitude and significance, across all the specifications. Note that column 6 is equivalent to column 1 in Table 5 in the main text.

Table 8 Nested models for local markers of firm productivity

Table 9 reports the main results with different threshold for the identification of a LLMA as UA. As in the main text, we first rely on the distribution of population between LLMAs in year 2001 but apply different cut-off values, above the \(75{\mathrm{th}}\), the \(95{\mathrm{th}}\) or the \(99{\mathrm{th}}\) percentile; moreover, we also apply the same arbitrary cut-off as in Giacinto et al. (2014), above 500 thousand residents.

Table 9 Local markers of firm TFP, changing UA definition

Table 10 controls for unobservable determinant of firm performance at the regional level by including a full set of dummy; results are largely confirmed, especially regarding the primary role of credit access as well as innovation and agglomeration economies.

Table 10 Local markers of firm TFP, controlling for regional unobservable

Finally, in Table 11 we test our results to the exclusion of firms with a capital intensity or value added per worker above/below the 99th/1st percentile of the industry-year distribution. Extreme values may be due to miss reporting but may also be induced by management strategies (e.g. by moving all the real estate assets to a separate legal entity) that may affect the reliability of production function estimates. Results largely confirm our main findings.

Table 11 Local markers of firm TFP, excluding outliers

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Manzocchi, S., Quintieri, B. & Santoni, G. Local manufacturing productivity markers: an empirical study of the Italian counties. Ann Reg Sci 59, 255–279 (2017).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:

JEL Classification

  • R11
  • R15
  • L60